
Aspect-based sentiment analysis (ABSA) is a fine-grained and diverse task in natural language processing. Existing deep learning models for ABSA face the challenge of balancing the demand for finer granularity in sentiment analysis with the scarcity of training corpora for such granularity. To address this issue, we propose an enhanced BERT-based model for multi-dimensional aspect target semantic learning. Our model leverages BERT's pre-training and fine-tuning mechanisms, enabling it to capture rich semantic feature parameters. In addition, we propose a complex semantic enhancement mechanism for aspect targets to enrich and optimize fine-grained training corpora. Third, we combine the aspect recognition enhancement mechanism with a CRF model to achieve more robust and accurate entity recognition for aspect targets. Furthermore, we propose an adaptive local attention mechanism learning model to focus on sentiment elements around rich aspect target semantics. Finally, to address the varying contributions of each task in the joint training mechanism, we carefully optimize this training approach, allowing for a mutually beneficial training of multiple tasks. Experimental results on four Chinese and five English datasets demonstrate that our proposed mechanisms and methods effectively improve ABSA models, surpassing some of the latest models in multi-task and single-task scenarios.
Citation: Quan Zhu, Xiaoyin Wang, Xuan Liu, Wanru Du, Xingxing Ding. Multi-task learning for aspect level semantic classification combining complex aspect target semantic enhancement and adaptive local focus[J]. Mathematical Biosciences and Engineering, 2023, 20(10): 18566-18591. doi: 10.3934/mbe.2023824
[1] | Debao Yan . Existence results of fractional differential equations with nonlocal double-integral boundary conditions. Mathematical Biosciences and Engineering, 2023, 20(3): 4437-4454. doi: 10.3934/mbe.2023206 |
[2] | Abdon Atangana, Jyoti Mishra . Analysis of nonlinear ordinary differential equations with the generalized Mittag-Leffler kernel. Mathematical Biosciences and Engineering, 2023, 20(11): 19763-19780. doi: 10.3934/mbe.2023875 |
[3] | Allaberen Ashyralyev, Evren Hincal, Bilgen Kaymakamzade . Crank-Nicholson difference scheme for the system of nonlinear parabolic equations observing epidemic models with general nonlinear incidence rate. Mathematical Biosciences and Engineering, 2021, 18(6): 8883-8904. doi: 10.3934/mbe.2021438 |
[4] | Sebastian Builes, Jhoana P. Romero-Leiton, Leon A. Valencia . Deterministic, stochastic and fractional mathematical approaches applied to AMR. Mathematical Biosciences and Engineering, 2025, 22(2): 389-414. doi: 10.3934/mbe.2025015 |
[5] | Hardik Joshi, Brajesh Kumar Jha, Mehmet Yavuz . Modelling and analysis of fractional-order vaccination model for control of COVID-19 outbreak using real data. Mathematical Biosciences and Engineering, 2023, 20(1): 213-240. doi: 10.3934/mbe.2023010 |
[6] | Barbara Łupińska, Ewa Schmeidel . Analysis of some Katugampola fractional differential equations with fractional boundary conditions. Mathematical Biosciences and Engineering, 2021, 18(6): 7269-7279. doi: 10.3934/mbe.2021359 |
[7] | Jian Huang, Zhongdi Cen, Aimin Xu . An efficient numerical method for a time-fractional telegraph equation. Mathematical Biosciences and Engineering, 2022, 19(5): 4672-4689. doi: 10.3934/mbe.2022217 |
[8] | Yingying Xu, Chunhe Song, Chu Wang . Few-shot bearing fault detection based on multi-dimensional convolution and attention mechanism. Mathematical Biosciences and Engineering, 2024, 21(4): 4886-4907. doi: 10.3934/mbe.2024216 |
[9] | H. M. Srivastava, Khaled M. Saad, J. F. Gómez-Aguilar, Abdulrhman A. Almadiy . Some new mathematical models of the fractional-order system of human immune against IAV infection. Mathematical Biosciences and Engineering, 2020, 17(5): 4942-4969. doi: 10.3934/mbe.2020268 |
[10] | Guodong Li, Ying Zhang, Yajuan Guan, Wenjie Li . Stability analysis of multi-point boundary conditions for fractional differential equation with non-instantaneous integral impulse. Mathematical Biosciences and Engineering, 2023, 20(4): 7020-7041. doi: 10.3934/mbe.2023303 |
Aspect-based sentiment analysis (ABSA) is a fine-grained and diverse task in natural language processing. Existing deep learning models for ABSA face the challenge of balancing the demand for finer granularity in sentiment analysis with the scarcity of training corpora for such granularity. To address this issue, we propose an enhanced BERT-based model for multi-dimensional aspect target semantic learning. Our model leverages BERT's pre-training and fine-tuning mechanisms, enabling it to capture rich semantic feature parameters. In addition, we propose a complex semantic enhancement mechanism for aspect targets to enrich and optimize fine-grained training corpora. Third, we combine the aspect recognition enhancement mechanism with a CRF model to achieve more robust and accurate entity recognition for aspect targets. Furthermore, we propose an adaptive local attention mechanism learning model to focus on sentiment elements around rich aspect target semantics. Finally, to address the varying contributions of each task in the joint training mechanism, we carefully optimize this training approach, allowing for a mutually beneficial training of multiple tasks. Experimental results on four Chinese and five English datasets demonstrate that our proposed mechanisms and methods effectively improve ABSA models, surpassing some of the latest models in multi-task and single-task scenarios.
Fractional calculus is a main branch of mathematics that can be considered as the generalisation of integration and differentiation to arbitrary orders. This hypothesis begins with the assumptions of L. Euler (1730) and G. W. Leibniz (1695). Fractional differential equations (FDEs) have lately gained attention and publicity due to their realistic and accurate computations [1,2,3,4,5,6,7]. There are various types of fractional derivatives, including Riemann–Liouville, Caputo, Grü nwald–Letnikov, Weyl, Marchaud, and Atangana. This topic's history can be found in [8,9,10,11]. Undoubtedly, fractional calculus applies to mathematical models of different phenomena, sometimes more effectively than ordinary calculus [12,13]. As a result, it can illustrate a wide range of dynamical and engineering models with greater precision. Applications have been developed and investigated in a variety of scientific and engineering fields over the last few decades, including bioengineering [14], mechanics [15], optics [16], physics [17], mathematical biology, electrical power systems [18,19,20] and signal processing [21,22,23].
One of the definitions of fractional derivatives is Caputo-Fabrizo, which adds a new dimension in the study of FDEs. The new derivative's feature is that it has a nonsingular kernel, which is made from a combination of an ordinary derivative with an exponential function, but it has the same supplementary motivating properties with various scales as in the Riemann-Liouville fractional derivatives and Caputo. The Caputo-Fabrizio fractional derivative has been used to solve real-world problems in numerous areas of mathematical modelling for example, numerical solutions for groundwater pollution, the movement of waves on the surface of shallow water modelling [24], RLC circuit modelling [25], and heat transfer modelling [26,27] were discussed.
Rach (1987), Bellomo and Sarafyan (1987) first compared the Adomian Decomposition method (ADM) [28,29,30,31,32] to the Picard method on a variety of examples. These methods have many benefits: they effectively work with various types of linear and nonlinear equations and also provide an analytic solution for all of these equations with no linearization or discretization. These methods are more realistic compared with other numerical methods as each technique is used to solve a specific type of equations, on the other hand ADM and Picard are useful for many types of equations. In the numerical examples provided, we compare ADM and Picard solutions of multidimentional fractional order equations with Caputo-Fabrizio.
The fractional derivative of Caputo-Fabrizio for the function x(t) is defined as [33]
CFDα0x(t)=B(α)1−α∫t0dds(x(s)) e−α1−α(t−s)ds, | (1.1) |
and its corresponding fractional integral is
CFIαx(t)=1−αB(α)x(t)+αB(α)∫t0x (s)ds, 0<α<1, | (1.2) |
where x(t) be continuous and differentiable on [0, T]. Also, in the above definition, the function B(α)>0 is a normalized function which satisfy the condition B(0)=B(1)=0. The relation between the Caputo–Fabrizio fractional derivate and its corresponding integral is given by
(CFIα0)(CFDα0f(t))=f(t)−f(a). | (1.3) |
In this section, we will introduce a multidimentional FDE subject to the initial condition. Let α∈(0,1], 0<α1<α2<...,αm<1, and m is integer real number,
CFDx=f(t,x,CFDα1x,CFDα2x,...,CFDαmx,) ,x(0)=c0, | (2.1) |
where x=x(t),t∈J=[0,T],T∈R+,x∈C(J).
To facilitate the equation and make it easy for the calculation, we let x(t)=c0+X(t) so Eq (2.1) can be witten as
CFDαX=f(t,c0+X,CFDα1X,CFDα2X,...,CFDαmX), X(0)=0. | (2.2) |
the algorithm depends on converting the initial condition from a constant c0 to 0.
Let CFDαX=y(t) then X=CFIαy, so we have
CFDαiX= CFIα−αi CFDαX= CFIα−αiy, i=1,2,...,m. | (2.3) |
Substituting in Eq (2.2) we obtain
y=f(t,c0+ CFIαy, CFIα−α1y,..., CFIα−αmy). | (2.4) |
Assume f satisfies Lipschtiz condition with Lipschtiz constant L given by,
|f(t,y0,y1,...,ym)|−|f(t,z0,z1,...,zm)|≤Lm∑i=0|yi−zi|, | (2.5) |
which implies
|f(t,c0+CFIαy,CFIα−α1y,..,CFIα−αmy)−f(t,c0+CFIαz,CFIα−α1z,..,CFIα−αmz)|≤Lm∑i=0| CFIα−αiy− CFIα−αiz|. | (2.6) |
The solution algorithm of Eq (2.4) using ADM is,
y0(t)=a(t)yn+1(t)=An(t), j⩾0. | (2.7) |
where a(t) pocesses all free terms in Eq (2.4) and An are the Adomian polynomials of the nonlinear term which takes the form [34]
An=f(Sn)−n−1∑i=0Ai, | (2.8) |
where f(Sn)=∑ni=0Ai. Later, this accelerated formula of Adomian polynomial will be used in convergence analysis and error estimation. The solution of Eq (2.4) can be written in the form,
y(t)=∞∑i=0yi(t). | (2.9) |
lastly, the solution of the Eq (2.4) takes the form
x(t)=c0+X(t)=c0+ CFIαy(t). | (2.10) |
At which we convert the parameter to the initial form y to x in Eq (2.10), so we have the solution of the original Eq (2.1).
Define a mapping F:E→E where E=(C[J],‖⋅‖) is a Banach space of all continuous functions on J with the norm ‖x‖= max.
Theorem 3.1. Equation (2.4) has a unique solution whenever 0 < \phi < 1 where \phi = L\left(\sum_{i = 0}^{m}\frac{\left[ \left(\alpha-\alpha _{i}\right) \left(T-1\right) \right] +1}{B\left(\alpha -\alpha_{i}\right) }\right) .
Proof. First, we define the mapping F:E\rightarrow E as
\begin{equation*} Fy = f(t,c_{0}+\text{ }^{CF}I^{\alpha }y,\text{ }^{CF}I^{\alpha -\alpha _{1}}y,...,\text{ }^{CF}I^{\alpha -\alpha _{m}}y). \end{equation*} |
Let y and z\in E are two different solutions of Eq (2.4). Then
\begin{equation*} Fy-Fz = f(t,c_{0}+^{CF}I^{\alpha }y,^{CF}I^{\alpha -\alpha _{1}}y,..,^{CF}I^{\alpha -\alpha _{m}}y)-f(t,c_{0}+^{CF}I^{\alpha }z,^{CF}I^{\alpha -\alpha _{1}}z,...,^{CF}I^{\alpha -\alpha _{m}}z) \end{equation*} |
which implies that
\begin{eqnarray*} \left\vert Fy-Fz\right\vert & = &\left\vert f(t,c_{0}+\text{ }^{CF}I^{\alpha }y,\text{ }^{CF}I^{\alpha -\alpha _{1}}y,...,\text{ }^{CF}I^{\alpha -\alpha _{m}}y)\right. \\ &&-\left. f(t,c_{0}+\text{ }^{CF}I^{\alpha }z,\text{ }^{CF}I^{\alpha -\alpha _{1}}z,...,\text{ }^{CF}I^{\alpha -\alpha _{m}}z)\right\vert \\ &\leq &L\sum\limits_{i = 0}^{m}\left\vert \text{ }^{CF}I^{\alpha -\alpha _{i}}y-\text{ }^{CF}I^{\alpha -\alpha _{i}}z\right\vert \\ &\leq &L\sum\limits_{i = 0}^{m}\left\vert \frac{1-\left( \alpha -\alpha _{i}\right) }{ B\left( \alpha -\alpha _{i}\right) }\left( y-z\right) +\frac{\alpha -\alpha _{i}}{B\left( \alpha -\alpha _{i}\right) }\int_{0}^{t}\left( y-z\right) ds\right\vert \\ \left\Vert Fy-Fz\right\Vert &\leq &L\sum\limits_{i = 0}^{m}\frac{1-\left( \alpha -\alpha _{i}\right) }{B\left( \alpha -\alpha _{i}\right) }\underset{ t\epsilon J}{\max }\left\vert y-z\right\vert +\frac{\alpha -\alpha _{i}}{ B\left( \alpha -\alpha _{i}\right) }\underset{t\epsilon J}{\max }\left\vert y-z\right\vert \int_{0}^{t}ds \\ &\leq &L\sum\limits_{i = 0}^{m}\frac{1-\left( \alpha -\alpha _{i}\right) }{B\left( \alpha -\alpha _{i}\right) }\left\Vert y-z\right\Vert +\frac{\alpha -\alpha _{i}}{B\left( \alpha -\alpha _{i}\right) }\left\Vert y-z\right\Vert T \\ &\leq &L\left\Vert y-z\right\Vert \left( \sum\limits_{i = 0}^{m}\frac{1-\left( \alpha -\alpha _{i}\right) }{B\left( \alpha -\alpha _{i}\right) }+\frac{\alpha -\alpha _{i}}{B\left( \alpha -\alpha _{i}\right) }T\right) \\ &\leq &L\left\Vert y-z\right\Vert \left( \sum\limits_{i = 0}^{m}\frac{\left[ \left( \alpha -\alpha _{i}\right) \left( T-1\right) \right] +1}{B\left( \alpha -\alpha _{i}\right) }\right) \\ &\leq &\phi \left\Vert y-z\right\Vert . \end{eqnarray*} |
under the condition 0 < \phi < 1, the mapping F is contraction and hence there exists a unique solution y\in C\left[ J\right] for the problem Eq (2.4) and this completes the proof.
Theorem 3.2. The series solution of the problem Eq (2.4)converges if \left\vert y_{1}\left(t\right) \right\vert < c and c isfinite.
Proof. Define a sequence \left\{ S_{p}\right\} such that S_{p} = \sum_{i = 0}^{p}y_{i}\left(t\right) is the sequence of partial sums from the series solution \sum_{i = 0}^{\infty }y_{i}\left(t\right), we have
\begin{equation*} f(t,c_{0}+\text{ }^{CF}I^{\alpha }y,\text{ }^{CF}I^{\alpha -\alpha _{1}}y,...,\text{ }^{CF}I^{\alpha -\alpha _{m}}y) = \sum\limits_{i = 0}^{\infty }A_{i}, \end{equation*} |
So
\begin{equation*} f(t,c_{0}+\text{ }^{CF}I^{\alpha }S_{p},\text{ }^{CF}I^{\alpha -\alpha _{1}}S_{p},...,\text{ }^{CF}I^{\alpha -\alpha _{m}}S_{p}) = \sum\limits_{i = 0}^{p}A_{i}, \end{equation*} |
From Eq (2.7) we have
\begin{equation*} \sum\limits_{i = 0}^{\infty }y_{i}\left( t\right) = a\left( t\right) +\sum\limits_{i = 0}^{\infty }A_{i-1} \end{equation*} |
let S_{p}, S_{q} be two arbitrary sums with p\geqslant q . Now, we are going to prove that \left\{ S_{p}\right\} is a Caushy sequence in this Banach space. We have
\begin{eqnarray*} S_{p} & = &\sum\limits_{i = 0}^{p}y_{i}\left( t\right) = a\left( t\right) +\sum\limits_{i = 0}^{p}A_{i-1,} \\ S_{q} & = &\sum\limits_{i = 0}^{q}y_{i}\left( t\right) = a\left( t\right) +\sum\limits_{i = 0}^{q}A_{i-1.} \end{eqnarray*} |
\begin{eqnarray*} S_{p}-S_{q} & = &\sum\limits_{i = 0}^{p}A_{i-1}-\sum\limits_{i = 0}^{q}A_{i-1} = \sum\limits_{i = q+1}^{p}A_{i-1} = \sum\limits_{i = q}^{p-1}A_{i-1} \\ & = &f(t,c_{0}+\text{ }^{CF}I^{\alpha }S_{p-1},\text{ }^{CF}I^{\alpha -\alpha _{1}}S_{p-1},...,\text{ }^{CF}I^{\alpha -\alpha _{m}}S_{p-1})- \\ &&f(t,c_{0}+\text{ }^{CF}I^{\alpha }S_{q-1},\text{ }^{CF}I^{\alpha -\alpha _{1}}S_{q-1},...,\text{ }^{CF}I^{\alpha -\alpha _{m}}S_{q-1}) \end{eqnarray*} |
\begin{eqnarray*} \left\vert S_{p}-S_{q}\right\vert & = &\left\vert f(t,c_{0}+\text{ } ^{CF}I^{\alpha }S_{p-1},\text{ }^{CF}I^{\alpha -\alpha _{1}}S_{p-1},..., \text{ }^{CF}I^{\alpha -\alpha _{m}}S_{p-1})-\right. \\ &&\left. f(t,c_{0}+\text{ }^{CF}I^{\alpha }S_{q-1},\text{ }^{CF}I^{\alpha -\alpha _{1}}S_{q-1},...,\text{ }^{CF}I^{\alpha -\alpha _{m}}S_{q-1})\right\vert \\ &\leq &L\sum\limits_{i = 0}^{m}\left\vert \text{ }^{CF}I^{\alpha -\alpha _{i}}S_{p-1}- \text{ }^{CF}I^{\alpha -\alpha _{i}}S_{q-1}\right\vert \\ &\leq &L\sum\limits_{i = 0}^{m}\left\vert \frac{1-\left( \alpha -\alpha _{i}\right) }{ B\left( \alpha -\alpha _{i}\right) }\left( S_{p-1}-S_{q-1}\right) +\frac{ \alpha -\alpha _{i}}{B\left( \alpha -\alpha _{i}\right) }\int_{0}^{t}\left( S_{p-1}-S_{q-1}\right) ds\right\vert \\ &\leq &L\sum\limits_{i = 0}^{m}\frac{1-\left( \alpha -\alpha _{i}\right) }{B\left( \alpha -\alpha _{i}\right) }\left\vert S_{p-1}-S_{q-1}\right\vert +\frac{ \alpha -\alpha _{i}}{B\left( \alpha -\alpha _{i}\right) }\int_{0}^{t}\left \vert S_{p-1}-S_{q-1}\right\vert ds \end{eqnarray*} |
\begin{eqnarray*} \left\Vert S_{p}-S_{q}\right\Vert &\leq &L\sum\limits_{i = 0}^{m}\frac{1-\left( \alpha -\alpha _{i}\right) }{B\left( \alpha -\alpha _{i}\right) }\underset{ t\epsilon J}{\max }\left\vert S_{p-1}-S_{q-1}\right\vert +\frac{\alpha -\alpha _{i}}{B\left( \alpha -\alpha _{i}\right) }\underset{t\epsilon J}{ \max }\left\vert S_{p-1}-S_{q-1}\right\vert \int_{0}^{t}ds \\ &\leq &L\left\Vert S_{p}-S_{q}\right\Vert \sum\limits_{i = 0}^{m}\left( \frac{ 1-\left( \alpha -\alpha _{i}\right) }{B\left( \alpha -\alpha _{i}\right) }+ \frac{\alpha -\alpha _{i}}{B\left( \alpha -\alpha _{i}\right) }T\right) \\ &\leq &L\left\Vert S_{p}-S_{q}\right\Vert \left( \sum\limits_{i = 0}^{m}\frac{\left[ \left( \alpha -\alpha _{i}\right) \left( T-1\right) \right] +1}{B\left( \alpha -\alpha _{i}\right) }\right) \\ &\leq &\phi \left\Vert S_{p}-S_{q}\right\Vert \end{eqnarray*} |
let p = q+1 then,
\begin{equation*} \left\Vert S_{q+1}-S_{q}\right\Vert \leq \phi \left\Vert S_{q}-S_{q-1}\right\Vert \leq \phi ^{2}\left\Vert S_{q-1}-S_{q-2}\right\Vert \leq ...\leq \phi ^{q}\left\Vert S_{1}-S_{0}\right\Vert \end{equation*} |
From the triangle inequality we have
\begin{eqnarray*} \left\Vert S_{p}-S_{q}\right\Vert &\leq &\left\Vert S_{q+1}-S_{q}\right\Vert +\left\Vert S_{q+2}-S_{q+1}\right\Vert +...\left\Vert S_{p}-S_{p-1}\right\Vert \\ &\leq &\left[ \phi ^{q}+\phi ^{q+1}+...+\phi ^{p-1}\right] \left\Vert S_{1}-S_{0}\right\Vert \\ &\leq &\phi ^{q}\left[ 1+\phi +...+\phi ^{p-q+1}\right] \left\Vert S_{1}-S_{0}\right\Vert \\ &\leq &\phi ^{q}\left[ \frac{1-\phi ^{p-q}}{1-\phi }\right] \left\Vert y_{1}\left( t\right) \right\Vert \end{eqnarray*} |
Since 0 < \phi < 1, p\geqslant q then \left(1-\phi ^{p-q}\right) \leq 1 . Consequently
\begin{equation} \left\Vert S_{p}-S_{q}\right\Vert \leq \frac{\phi ^{q}}{1-\phi }\left\Vert y_{1}\left( t\right) \right\Vert \leq \frac{\phi ^{q}}{1-\phi }\underset{ \forall t\epsilon J}{\max }\left\vert y_{1}\left( t\right) \right\vert \end{equation} | (3.1) |
but \left\vert y_{1}\left(t\right) \right\vert < \infty and as q\rightarrow \infty then, \left\Vert S_{p}-S_{q}\right\Vert \rightarrow 0 and hence, \left\{ S_{p}\right\} is a Caushy sequence in this Banach space then the proof is complete.
Theorem 3.3. The maximum absolute truncated error Eq (2.4)is estimated to be \underset{t\epsilon J}{\max }\left\vert y\left(t\right)-\sum_{i = 0}^{q}y_{i}\left(t\right) \right\vert \leq \frac{\phi ^{q}}{1-\phi }\underset{t\epsilon J}{\max }\left\vert y_{1}\left(t\right) \right\vert
Proof. From the convergence theorm inequality (Eq 3.1) we have
\begin{equation*} \left\Vert S_{p}-S_{q}\right\Vert \leq \frac{\phi ^{q}}{1-\phi }\underset{ t\epsilon J}{\max }\left\vert y_{1}\left( t\right) \right\vert \end{equation*} |
but, S_{p} = \sum_{i = 0}^{p}y_{i}\left(t\right) as p\rightarrow \infty then, S_{p}\rightarrow y\left(t\right) so,
\begin{equation*} \left\Vert y\left( t\right) -S_{q}\right\Vert \leq \frac{\phi ^{q}}{1-\phi } \underset{t\epsilon J}{\max }\left\vert y_{1}\left( t\right) \right\vert \end{equation*} |
so, the maximum absolute truncated error in the interval J is,
\begin{equation} \underset{t\epsilon J}{\max }\left\vert y\left( t\right) -\sum\limits_{i = 0}^{q}y_{i}\left( t\right) \right\vert \leq \frac{\phi ^{q}}{1-\phi }\underset{t\epsilon J}{\max }\left\vert y_{1}\left( t\right) \right\vert \end{equation} | (3.2) |
and this completes the proof.
In this part, we introduce several numerical examples with unkown exact solution and we will use inequality (Eq 3.2) to estimate the maximum absolute truncated error.
Example 4.1. Application of linear FDE
\begin{equation} ^{CF}Dx\left( t\right) +2a^{CF}D^{1/2}x\left( t\right) +bx\left( t\right) = 0, { \ \ \ \ \ \ \ }x\left( 0\right) = 1. \end{equation} | (4.1) |
A Basset problem in fluid dynamics is a classical problem which is used to study the unsteady movement of an accelerating particle in a viscous fluid under the action of the gravity [36]
Set
\begin{equation*} X\left( t\right) = x\left( t\right) -1 \end{equation*} |
Equation (4.1) will be
\begin{equation} ^{CF}DX\left( t\right) +2a^{CF}D^{1/2}X\left( t\right) +bX\left( t\right) = 0, { \ \ \ \ \ \ \ }X\left( 0\right) = 0. \end{equation} | (4.2) |
Appling Eq (2.3) to Eq (4.2), and using initial condition, also we take a = 1, b = 1/2,
\begin{equation} y = -\frac{1}{2}-2I^{1/2}y-\frac{1}{2}I\text{ }y \end{equation} | (4.3) |
Appling ADM to Eq (4.3), we find the solution algorithm become
\begin{eqnarray} y_{0}\left( t\right) & = &-\frac{1}{2}, \\ y_{i}\left( t\right) & = &-2\text{ }^{CF}I^{1/2}y_{i-1}-\frac{1}{2}\text{ } ^{CF}I\text{ }y_{i-1},\ \ \ \ \ i\geq 1. \end{eqnarray} | (4.4) |
Appling Picard solution to Eq (4.2), we find the solution algorithm become
\begin{eqnarray} y_{0}\left( t\right) & = &-\frac{1}{2}, \\ y_{i}\left( t\right) & = &-\frac{1}{2}-2\text{ }^{CF}I^{1/2}y_{i-1}-\frac{1}{2} \text{ }^{CF}I\text{ }y_{i-1},\ \ \ \ \ i\geq 1. \end{eqnarray} | (4.5) |
From Eq (4.4), the solution using ADM is given by y\left(t\right) = \underset{q\rightarrow \infty }{Lim}{_{i = 0}^{q} y_{i}} \left(t\right) while from Eq (4.5), the solution using Picard technique is given by y\left(t\right) = \; \underset{i\rightarrow \infty }{ Lim} \; y_{i}\left(t\right) . Lately, the solution of the original problem Eq (4.2), is
\begin{equation*} x\left( t\right) = 1+\text{ }^{CF}I\text{ }y\left( t\right) . \end{equation*} |
One the same processor (q = 20), the time consumed using ADM is 0.037 seconds, while the time consumed using Picard is 7.955 seconds.
Figure 1 gives a comparison between ADM and Picard solution of Ex. 4.1.
Example 4.2. Consider the following nonlinear FDE [35]
\begin{eqnarray} ^{CF}D^{1/2}x & = &\frac{8t^{3/2}}{3\sqrt{\pi }}-\frac{t^{7/4}}{4\Gamma \left( \frac{11}{4}\right) }-\frac{t^{4}}{4}+\frac{1}{8}\text{ }^{CF}D^{1/4}x+\frac{ 1}{4}x^{2},\text{ } \\ x\left( 0\right) & = &0. \end{eqnarray} | (4.6) |
Appling Eq (2.3) to Eq (4.6), and using initial condition,
\begin{equation} y = \frac{8t^{3/2}}{3\sqrt{\pi }}-\frac{t^{7/4}}{4\Gamma \left( \frac{11}{4} \right) }-\frac{t^{4}}{4}+\frac{1}{8}\text{ }^{CF}I^{1/4}y+\frac{1}{4}\left( ^{CF}I^{1/2}y\right) ^{2}. \end{equation} | (4.7) |
Appling ADM to Eq (4.7), we find the solution algorithm will be become
\begin{eqnarray} y_{0}\left( t\right) & = &\frac{8t^{3/2}}{3\sqrt{\pi }}-\frac{t^{7/4}}{4\Gamma \left( \frac{11}{4}\right) }-\frac{t^{4}}{4}, \\ y_{i}\left( t\right) & = &\frac{1}{8}\text{ }^{CF}I^{1/4}y_{i-1}+\frac{1}{4} \left( A_{i-1}\right) ,\ \ \ \ \ i\geq 1. \end{eqnarray} | (4.8) |
at which A _{\text{i}} are Adomian polynomial of the nonliner term \left(^{CF}I^{1/2}y\right) ^{2}.
Appling Picard solution to Eq (4.7), we find the the solution algorithm become
\begin{eqnarray} y_{0}\left( t\right) & = &\frac{8t^{3/2}}{3\sqrt{\pi }}-\frac{t^{7/4}}{4\Gamma \left( \frac{11}{4}\right) }-\frac{t^{4}}{4}, \\ y_{i}\left( t\right) & = &y_{0}\left( t\right) +\frac{1}{8}\text{ } ^{CF}I^{1/4}y_{i-1}+\frac{1}{4}\left( ^{CF}I^{1/2}y_{i-1}\right) ^{2},\ \ \ \ \ i\geq 1. \end{eqnarray} | (4.9) |
From Eq (4.8), the solution using ADM is given by y\left(t\right) = \underset{q\rightarrow \infty }{Lim}{_{i = 0}^{q}y_{i}} \left(t\right) while from Eq (4.9), the solution using Picard technique is given by y\left(t\right) = \underset{i\rightarrow \infty }{Lim} y_{i}\left(t\right) . Finally, the solution of the original problem Eq (4.7), is.
\begin{equation*} x\left( t\right) = \text{ }^{CF}I^{1/2}y. \end{equation*} |
One the same processor (q = 2), the time consumed using ADM is 65.13 seconds, while the time consumed using Picard is 544.787 seconds.
Table 1 showed the maximum absolute truncated error of of ADM solution (using Theorem 3.3) at different values of m (when t = 0:5; N = 2):
q | max. absolute error |
2 | 0.114548 |
5 | 0.099186 |
10 | 0.004363 |
Figure 2 gives a comparison between ADM and Picard solution of Ex. 4.2.
Example 4.3. Consider the following nonlinear FDE [35]
\begin{eqnarray} ^{CF}D^{\alpha }x & = &3t^{2}-\frac{128}{125\pi }t^{5}+\frac{1}{10}\left( ^{CF}D^{1/2}x\right) ^{2}, \\ x\left( 0\right) & = &0. \end{eqnarray} | (4.10) |
Appling Eq (2.3) to Eq (4.10), and using initial condition,
\begin{equation} y = 3t^{2}-\frac{128}{125\pi }t^{5}+\frac{1}{10}\left( ^{CF}I^{1/2}y\right) ^{2} \end{equation} | (4.11) |
Appling ADM to Eq (4.11), we find the solution algorithm become
\begin{eqnarray} y_{0}\left( t\right) & = &3t^{2}-\frac{128}{125\pi }t^{5}, \\ y_{i}\left( t\right) & = &\frac{1}{10}\left( A_{i-1}\right) ,\ \ \ \ \ i\geq 1 \end{eqnarray} | (4.12) |
at which A _{\text{i}} are Adomian polynomial of the nonliner term \left(^{CF}I^{1/2}y\right) ^{2}.
Then appling Picard solution to Eq (4.11), we find the solution algorithm become
\begin{eqnarray} y_{0}\left( t\right) & = &3t^{2}-\frac{128}{125\pi }t^{5}, \\ y_{i}\left( t\right) & = &y_{0}\left( t\right) +\frac{1}{10}\left( ^{CF}I^{1/2}y_{i-1}\right) ^{2},\ \ \ \ \ i\geq 1. \end{eqnarray} | (4.13) |
From Eq (4.12), the solution using ADM is given by y\left(t\right) = \underset{q\rightarrow \infty }{Lim}{_{i = 0}^{q}y_{i}} \left(t\right) while from Eq (4.13), the solution is y\left(t\right) = \underset{i\rightarrow \infty }{Lim}y_{i}\left(t\right) . Finally, the solution of the original problem Eq (4.11), is
\begin{equation*} x\left( t\right) = ^{CF}Iy\left( t\right) . \end{equation*} |
One the same processor (q = 4), the time consumed using ADM is 2.09 seconds, while the time consumed using Picard is 44.725 seconds.
Table 2 showed the maximum absolute truncated error of of ADM solution (using Theorem 3.3) at different values of m (when t = 0:5; N = 4):
q | max. absolute error |
2 | 0.00222433 |
5 | 0.0000326908 |
10 | 2.88273*10 ^{-8} |
Figure 3 gives a comparison between ADM and Picard solution of Ex. 4.3 with \alpha = 1 .
Example 4.4. Consider the following nonlinear FDE [35]
\begin{eqnarray} ^{CF}D^{\alpha }x & = &t^{2}+\frac{1}{2}\text{ }^{CF}D^{\alpha _{1}}x+\frac{1}{ 4}\text{ }^{CF}D^{\alpha _{2}}x+\frac{1}{6}\text{ }^{CF}D^{\alpha 3}x+\frac{1 }{8}x^{4}, \\ x\left( 0\right) & = &0. \end{eqnarray} | (4.14) |
Appling Eq (2.3) to Eq (4.10), and using initial condition,
\begin{equation} y = t^{2}+\frac{1}{2}\left( ^{CF}I^{\alpha -\alpha _{1}}y\right) +\frac{1}{4} \left( ^{CF}I^{\alpha -\alpha _{2}}y\right) +\frac{1}{6}\left( ^{CF}I^{\alpha -\alpha 3}y\right) +\frac{1}{8}\left( ^{CF}I^{\alpha }y\right) ^{4}, \end{equation} | (4.15) |
Appling ADM to Eq (4.15), we find the solution algorithm become
\begin{eqnarray} y_{0}\left( t\right) & = &t^{2}, \\ y_{i}\left( t\right) & = &\frac{1}{2}\left( ^{CF}I^{\alpha -\alpha _{1}}y\right) +\frac{1}{4}\left( ^{CF}I^{\alpha -\alpha _{2}}y\right) +\frac{ 1}{6}\left( ^{CF}I^{\alpha -\alpha 3}y\right) +\frac{1}{8}A_{i-1},{ \ \ }i\geq 1 \end{eqnarray} | (4.16) |
where A _{\text{i}} are Adomian polynomial of the nonliner term \left(^{CF}I^{\alpha }y\right) ^{4}.
Then appling Picard solution to Eq (4.15), we find the solution algorithm become
y_{0}\left( t\right) = t^{2}, \\ y_{i}\left( t\right) = t^{2}+\frac{1}{2}\left( ^{CF}I^{\alpha -\alpha _{1}}y_{i-1}\right) +\frac{1}{4}\left( ^{CF}I^{\alpha -\alpha _{2}}y_{i-1}\right) \\+\frac{1}{6}\left( ^{CF}I^{\alpha -\alpha 3}y_{i-1}\right) +\frac{1}{8}\left( ^{CF}I^{\alpha }y_{i-1}\right) ^{4}\ \ \ \ \ i\geq 1. | (4.17) |
From Eq (4.16), the solution using ADM is given by y\left(t\right) = \underset{q\rightarrow \infty }{Lim}{_{i = 0}^{q}y_{i}} \left(t\right) while from Eq (4.17), the solution using Picard technique is y\left(t\right) = \underset{i\rightarrow \infty }{Lim} y_{i}\left(t\right) . Finally, the solution of the original problem Eq (4.14), is
\begin{equation*} x\left( t\right) = ^{CF}I^{\alpha }y\left( t\right) . \end{equation*} |
One the same processor (q = 3), the time consumed using ADM is 0.437 seconds, while the time consumed using Picard is (16.816) seconds. Figure 4 shows a comparison between ADM and Picard solution of Ex. 4.4 at \; \alpha = 0.7, \; \alpha _{1} = 0.1, \alpha _{2} = 0.3, \alpha _{3} = 0.5.
The Caputo-Fabrizo fractional deivative has a nonsingular kernel, and consequently, this definition is appropriate in solving nonlinear multidimensional FDE [37,38]. Since the selected numerical problems have an unkown exact solution, the formula (3.2) can be used to estimate the maximum absolute truncated error. By comparing the time taken on the same processor (i7-2670QM), it was found that the time consumed by ADM is much smaller compared with the Picard technique. Furthermore Picard gives a more accurate solution than ADM at the same interval with the same number of terms.
The authors declare there is no conflict of interest.
[1] |
B. Pang, L. Lee, Opinion mining and sentiment analysis, Trends Inf. Retr., 2 (2008), 1–135. https://doi.org/10.1561/1500000011 doi: 10.1561/1500000011
![]() |
[2] |
G. Vinodhini, R. Chandrasekaran, Sentiment analysis and opinion mining: a survey, Int. J., 2 (2012), 282–292. https://doi.org/10.1016/j.nlp.2022.100003 doi: 10.1016/j.nlp.2022.100003
![]() |
[3] | M. Pontiki, D. Galanis, J. Pavlopoulos, H. Papageorgiou, S. Manandhar, SemEval-2014 Task 4: Aspect based sentiment analysis, in Association for Computational Linguistics, (2014), 27–35. https://doi.org/10.3115/v1/S14-2004 |
[4] | M. Pontiki, D. Galanis, H. Papageorgiou, S. Manandhar, I. Androutsopoulos, Semeval-2015 task 12: Aspect based sentiment analysis, in Association for Computational Linguistics, (2015), 486–495. https://doi.org/10.18653/v1/S15-2082 |
[5] | M. Pontiki, D. Galanis, H. Papageorgiou, I. Androutsopoulos, S. Manandhar, M. AL-Smadi, et al. Semeval-2016 task 5: Aspect based sentiment analysis, in Association for Computational Linguistics, (2016), 19–30. https://doi.org/10.18653/v1/S16-1002 |
[6] |
W. Zhang, X. Li, Y. Deng, L. Bing, W. Lam, A survey on aspect-based sentiment analysis: Tasks, methods, and challenges, IEEE Trans. Knowl. Data Eng., 2022. https://doi.org/10.1109/TKDE.2022.3230975 doi: 10.1109/TKDE.2022.3230975
![]() |
[7] | D. Tang, B. Qin, X. Feng, T. Liu, Effective LSTMs for target-dependent sentiment classification, preprint, arXiv: 151201100. |
[8] | M. Yang, W. Tu, J. Wang, F. Xu, X. Chen, Attention based LSTM for target dependent sentiment classification, in Proceedings of the AAAI conference on artificial intelligence, 2017. https://doi.org/10.1609/aaai.v31i1.11061 |
[9] | Q. Liu, Y. Huang, Q. Yang, H. Peng, J. Wang, An attention-aware long short-term memory-like spiking neural model for sentiment analysis, Int. J. Neural Syst., (2023), 2350037–2350037. https://doi.org/10.1142/s0129065723500375 |
[10] |
Y. Huang, Q. Liu, H. Peng, J. Wang, Q. Yang, D. Orellana-Martín, Sentiment classification using bidirectional LSTM-SNP model and attention mechanism, Expert Syst. Appl., 221 (2023), 119730. https://doi.org/10.1016/j.eswa.2023.119730 doi: 10.1016/j.eswa.2023.119730
![]() |
[11] |
Y. Huang, H. Peng, Q. Liu, Q. Yang, J. Wang, D. Orellana-Martín, et al., Attention-enabled gated spiking neural P model for aspect-level sentiment classification, Neural Network, 157 (2023), 437–443. https://doi.org/10.1016/j.neunet.2022.11.006 doi: 10.1016/j.neunet.2022.11.006
![]() |
[12] | Y. Kim, Convolutional neural networks for sentence classification, preprint, arXiv: 14085882. |
[13] | D. Tang, B. Qin, T. Liu, Aspect level sentiment classification with deep memory network, preprint, arXiv: 160508900. |
[14] | P. Lin, M. Yang, J. Lai. Deep mask memory network with semantic dependency and context moment for aspect level sentiment classification, in IJCAI, (2019), 5088–5094. https://doi.org/10.24963/ijcai.2019/707 |
[15] | A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, et al., Attention is all you need, in Advances in Neural Information Processing Systems, 30 (2017). https://doi.org/10.48550/arXiv.1706.03762 |
[16] | Z.-Y. Dou, Capturing user and product information for document level sentiment analysis with deep memory network, in Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, 2017. https://doi.org/10.18653/v1/D17-1054 |
[17] |
K. Chakraborty, S. Bhattacharyya, R. Bag, A survey of sentiment analysis from social media data, IEEE Trans. Comput. Soc. Syst., 7 (2020), 450–464. https://doi.org/10.1109/TCSS.2019.2956957 doi: 10.1109/TCSS.2019.2956957
![]() |
[18] |
X. Zhu, Y. Zhu, L. Zhang, Y. Chen, A BERT-based multi-semantic learning model with aspect-aware enhancement for aspect polarity classification, Appl. Intell., 53 (2023), 4609–4623. https://doi.org/10.1007/s10489-022-03702-1 doi: 10.1007/s10489-022-03702-1
![]() |
[19] | J. Devlin, M. W. Chang, K. Lee, K. Toutanova, Bert: Pre-training of deep bidirectional transformers for language understanding, preprint, arXiv: 181004805. |
[20] | N. Reimers, I. Gurevych, Sentence-bert: Sentence embeddings using siamese bert-networks, preprint, arXiv: 190810084. |
[21] |
L. Breiman, J. Friedman, C. J. Stone, R. A. Olshen, Classification and Regression Trees (CART), Biometrics, 1984 (1984). https://doi.org/10.2307/2530946 doi: 10.2307/2530946
![]() |
[22] |
N. S. Altman, An introduction to kernel and nearest-neighbor nonparametric regression, Am. Stat., 46 (1992), 175–185. https://doi.org/10.1080/00031305.1992.10475879 doi: 10.1080/00031305.1992.10475879
![]() |
[23] | I. Rish, An empirical study of the naive Bayes classifier, in IJCAI 2001 workshop on empirical methods in artificial intelligence, (2001), 41–46. https://doi.org/10.1109/CSCI46756.2018.00065 |
[24] | D. W. Hosmer Jr, S. Lemeshow, R. X. Sturdivant, Applied Logistic Regression, John Wiley & Sons, 2013. https://doi.org/10.1002/9781118548387 |
[25] | C. Cortes, V. Vapnik, Support-vector networks, Mach. Learn., 20 (1995), 273–297. |
[26] |
L. Breiman, Random forests, Mach. Learn., 45 (2001), 5–32. https://doi.org/10.1023/A:1022627411411 doi: 10.1023/A:1022627411411
![]() |
[27] | N. S. Joshi, S. A. Itkat, A survey on feature level sentiment analysis, Int. J. Comput. Sci. Inf. Technol., 5 (2014), 5422–5425. |
[28] |
E. Cambria, B. White, Jumping NLP curves: A review of natural language processing research, IEEE Comput. Intell. Mag., 9 (2014), 48–57. https://doi.org/10.1109/MCI.2014.2307227 doi: 10.1109/MCI.2014.2307227
![]() |
[29] |
B. Zhang, X. Fu, C. Luo, Y. Ye, X. Li, L. Jing, Cross-domain aspect-based sentiment classification by exploiting domain-invariant semantic-primary feature, IEEE Trans. Affect. Comput., 2023 (2023), forthcoming. https://doi.org/10.1109/TAFFC.2023.3239540 doi: 10.1109/TAFFC.2023.3239540
![]() |
[30] |
H. Huang, B. Zhang, L. Jing, X. Fu, X. Chen, J. Shi, Logic tensor network with massive learned knowledge for aspect-based sentiment analysis, Knowl. Based Syst., 257 (2022), 109943. https://doi.org/10.1016/j.knosys.2022.109943 doi: 10.1016/j.knosys.2022.109943
![]() |
[31] |
X. Mei, Y. Zhou, C. Zhu, M. Wu, M. Li, S. Pan, A disentangled linguistic graph model for explainable aspect-based sentiment analysis, Knowl. Based Syst, 260 (2023), 110150. https://doi.org/10.1016/j.knosys.2022.110150 doi: 10.1016/j.knosys.2022.110150
![]() |
[32] | B. Zhang, X. Huang, Z. Huang, H. Huang, B. Zhang, X. Fu, et al., Sentiment interpretable logic tensor network for aspect-term sentiment analysis, in Proceedings of the 29th International Conference on Computational Linguistics, (2022), 6705–6714. |
[33] | B. Xu, X. Wang, B. Yang, Z. Kang, Target embedding and position attention with lstm for aspect based sentiment analysis, in Proceedings of the 2020 5th International Conference on Mathematics and Artificial Intelligence, (2020), 93–97. https://doi.org/10.1145/3395260.3395280 |
[34] | Y. Ma, H. Peng, E. Cambria, Targeted aspect-based sentiment analysis via embedding commonsense knowledge into an attentive LSTM, in Proceedings of the AAAI conference on artificial intelligence, (2018). https://doi.org/10.1609/aaai.v32i1.12048 |
[35] | L. Bao, P. Lambert, T. Badia, Attention and lexicon regularized LSTM for aspect-based sentiment analysis, in Proceedings of the 57th annual meeting of the association for computational linguistics: student research workshop, (2019), 253–259. https://doi.org/10.18653/v1/P19-2035 |
[36] |
Y. Xing, C. Xiao, Y. Wu, Z. Ding, A convolutional neural network for aspect-level sentiment classification, Int. J. Pattern Recognit. Artif Intell., 33 (2019), 1959046. https://doi.org/10.18653/v1/2021.textgraphs-1.8 doi: 10.18653/v1/2021.textgraphs-1.8
![]() |
[37] |
X. Wang, F. Li, Z. Zhang, G. Xu, J. Zhang, X. Sun, A unified position-aware convolutional neural network for aspect based sentiment analysis, Neurocomputing, 450 (2021), 91–103. https://doi.org/10.1016/j.neucom.2021.03.092 doi: 10.1016/j.neucom.2021.03.092
![]() |
[38] |
C. Gan, L. Wang, Z. Zhang, Z. Wang, Sparse attention based separable dilated convolutional neural network for targeted sentiment analysis, Knowl. Based Syst., 188 (2020), 104827. https://doi.org/10.1016/j.knosys.2019.06.035 doi: 10.1016/j.knosys.2019.06.035
![]() |
[39] |
N. Zhao, H. Gao, X. Wen, H. Li, Combination of convolutional neural network and gated recurrent unit for aspect-based sentiment analysis, IEEE Access, 9 (2021), 15561–15569. https://doi.org/10.1109/ACCESS.2021.3052937 doi: 10.1109/ACCESS.2021.3052937
![]() |
[40] | Y. Tay, L. A. Tuan, S. C. Hui, Dyadic memory networks for aspect-based sentiment analysis, in Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, (2017), 107–116. https://doi.org/10.1145/3132847.3132936 |
[41] |
Y. Chen, T. Zhuang, K. Guo, Memory network with hierarchical multi-head attention for aspect-based sentiment analysis, Appl. Intell., 51 (2021), 4287–4304. https://doi.org/10.1007/s10489-020-02069-5 doi: 10.1007/s10489-020-02069-5
![]() |
[42] |
Y. Zhang, B. Xu, T. Zhao, Convolutional multi-head self-attention on memory for aspect sentiment classification, IEEE-CAA J. Automatica Sin., 7 (2020), 1038–1044. https://doi.org/10.1109/JAS.2020.1003243 doi: 10.1109/JAS.2020.1003243
![]() |
[43] | Y. Song, J. Wang, T. Jiang, Z. Liu, Y. Rao, Attentional encoder network for targeted sentiment classification, preprint, arXiv: 190209314. |
[44] |
H. Yang, B. Zeng, J. Yang, Y. Song, R. Xu, A multi-task learning model for chinese-oriented aspect polarity classification and aspect term extraction, Neurocomputing, 419 (2021), 344–356. https://doi.org/10.1016/j.neucom.2020.08.001 doi: 10.1016/j.neucom.2020.08.001
![]() |
[45] | A. Karimi, L. Rossi, A. Prati, Improving bert performance for aspect-based sentiment analysis, preprint, arXiv: 201011731. |
[46] | A. Karimi, L. Rossi, A. Prati, Adversarial training for aspect-based sentiment analysis with bert, in 2020 25th International conference on pattern recognition (ICPR), (2021), 8797–8803. https://doi.org/10.1109/ICPR48806.2021.9412167 |
[47] | H. Peng, Y. Ma, Y. Li, E. Cambria, Learning multi-grained aspect target sequence for Chinese sentiment analysis, Knowl. Based Syst., 148 (2018), 167–176. |
[48] | W. Che, Y. Zhao, H. Guo, Z. Su, T. Liu, Sentence compression for aspect-based sentiment analysis, IEEE-ACM Trans. Audio Speech Lang., 23 (2015), 2111–2124. |
[49] | L. Dong, F. Wei, C. Tan, D. Tang, M. Zhou, K. Xu, Adaptive recursive neural network for target-dependent twitter sentiment classification, in Proceedings of the 52nd annual meeting of the association for computational linguistics (volume 2: Short papers), (2014), 49–54. |
[50] | B. Wang, W. Lu, Learning latent opinions for aspect-level sentiment classification, in Proceedings of the AAAI Conference on Artificial Intelligence, 2018. |
[51] | H. T. Nguyen, M. Le Nguyen, Effective attention networks for aspect-level sentiment classification, in 2018 10th International Conference on Knowledge and Systems Engineering (KSE), (2018), 25–30. https://doi.org/10.1109/KSE.2018.8573324 |
[52] | D. P. Kingma, J. Ba, Adam: A method for stochastic optimization, preprint, arXiv: 14126980. |
[53] | Y. Wang, M. Huang, X. Zhu, L. Zhao, Attention-based LSTM for aspect-level sentiment classification, in Proceedings of the 2016 conference on empirical methods in natural language processing, (2016), 606–615. https://doi.org/10.18653/v1/D16-1058 |
[54] | D. Ma, S. Li, X. Zhang, H. Wang, Interactive attention networks for aspect-level sentiment classification, preprint, arXiv: 170900893. |
[55] | H. Peng, L. Xu, L. Bing, F. Huang, W. Lu, L. Si, Knowing what, how and why: A near complete solution for aspect-based sentiment analysis, in Proceedings of the AAAI conference on artificial intelligence, (2020), 8600–8607. https://doi.org/10.1609/aaai.v34i05.6383 |
[56] | W. Song, Z. Wen, Z. Xiao, S. C. Park, Semantics perception and refinement network for aspect-based sentiment analysis, Knowl. Based Syst., 214 (2021), 106755. |
[57] | L. Xu, L. Bing, W. Lu, F. Huang, Aspect sentiment classification with aspect-specific opinion spans, in Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), (2020), 3561–3567. https://doi.org/10.18653/v1/2020.emnlp-main.288 |
[58] |
Q. Xu, L. Zhu, T. Dai, C. Yan, Aspect-based sentiment classification with multi-attention network, Neurocomputing, 388 (2020), 135–143. https://doi.org/10.1016/j.neucom.2020.01.024 doi: 10.1016/j.neucom.2020.01.024
![]() |
[59] |
B. Huang, J. Zhang, J. Ju, R. Guo, H. Fujita, J. Liu, CRF-GCN: An effective syntactic dependency model for aspect-level sentiment analysis, Knowl. Based Syst., 260 (2023), 110125. https://doi.org/10.1016/j.knosys.2022.110125 doi: 10.1016/j.knosys.2022.110125
![]() |
[60] |
B. Huang, R. Guo, Y. Zhu, Z. Fang, G. Zeng, J. Liu, et al., Aspect-level sentiment analysis with aspect-specific context position information, Knowl. Based Syst., 243 (2022), 108473. https://doi.org/10.1016/j.knosys.2022.108473 doi: 10.1016/j.knosys.2022.108473
![]() |
1. | Eman A. A. Ziada, Salwa El-Morsy, Osama Moaaz, Sameh S. Askar, Ahmad M. Alshamrani, Monica Botros, Solution of the SIR epidemic model of arbitrary orders containing Caputo-Fabrizio, Atangana-Baleanu and Caputo derivatives, 2024, 9, 2473-6988, 18324, 10.3934/math.2024894 | |
2. | H. Salah, M. Anis, C. Cesarano, S. S. Askar, A. M. Alshamrani, E. M. Elabbasy, Fourth-order differential equations with neutral delay: Investigation of monotonic and oscillatory features, 2024, 9, 2473-6988, 34224, 10.3934/math.20241630 | |
3. | Said R. Grace, Gokula N. Chhatria, S. Kaleeswari, Yousef Alnafisah, Osama Moaaz, Forced-Perturbed Fractional Differential Equations of Higher Order: Asymptotic Properties of Non-Oscillatory Solutions, 2024, 9, 2504-3110, 6, 10.3390/fractalfract9010006 | |
4. | A.E. Matouk, Monica Botros, Hidden chaotic attractors and self-excited chaotic attractors in a novel circuit system via Grünwald–Letnikov, Caputo-Fabrizio and Atangana-Baleanu fractional operators, 2025, 116, 11100168, 525, 10.1016/j.aej.2024.12.064 | |
5. | Zahra Barati, Maryam Keshavarzi, Samaneh Mosaferi, Anatomical and micromorphological study of Phalaris (Poaceae) species in Iran, 2025, 68, 1588-4082, 9, 10.14232/abs.2024.1.9-15 |
q | max. absolute error |
2 | 0.114548 |
5 | 0.099186 |
10 | 0.004363 |
q | max. absolute error |
2 | 0.00222433 |
5 | 0.0000326908 |
10 | 2.88273*10 ^{-8} |