Domain Adaptation via Tree Kernel Based Maximum Mean Discrepancy for User Consumption Intention Identification

Abstract

Identifying user consumption intention from social media is of great interests to downstream applications. Since such task is domain-dependent, deep neural networks have been applied to learn transferable features for adapting models from a source domain to a target domain. A basic idea to solve this problem is reducing the distribution difference between the source domain and the target domain such that the transfer error can be bounded. However, the feature transferability drops dramatically in higher layers of deep neural networks with increasing domain discrepancy. Hence, previous work has to use a few target domain annotated data to train domain-specific layers. In this paper, we propose a deep transfer learning framework for consumption intention identification, to reduce the data bias and enhance the transferability in domain-specific layers. In our framework, the representation of the domain-specific layer is mapped to a reproducing kernel Hilbert space, where the mean embeddings of different domain distributions can be explicitly matched. By using an optimal tree kernel method for measuring the mean embedding matching, the domain discrepancy can be effectively reduced. The framework can learn transferable features in a completely unsupervised manner with statistical guarantees. Experimental results on five different domain datasets show that our approach dramatically outperforms state-of-the-art baselines, and it is general enough to be applied to more scenarios. The source code and datasets can be found at http://ir.hit.edu.cn/$\scriptsize{\sim}$xding/index\_english.htm.

Beyond Sentiment Classification: A Novel Approach for Utilizing Social Media Data for Business Intelligence

“…In this section, we present the experimental results along with a discussion on the results. We used the dataset by [22] to investigate the efficiency of our approach in classifying product reviews based on those that signify purchase intention and those that do not. However, because we are able to get only one benchmark dataset, we randomly divided the dataset into three parts.…”

“…Therefore, there is very little benchmark dataset for purchase intention mining from social media. On product reviews specifically, we are able to find one dataset by [22]. The dataset consists of 7,522 instances, divided as 6,016 for training, 752 for development and 754 for testing.…”

Learning Multi-Domain Adversarial Neural Networks for Text Classification

“…Since [37] used more training data (5600 instances), which is not a publicly available dataset, we cannot compare with it in this paper. 7) DACI [4]: Tree-LSTM is used as the basic text classification model, and tree kernel based MMD is used to transfer task-specific layer's parameters from the source domain to the target domain. This method is used in the conventional domain adaptation framework.…”

“…To evaluate the effectiveness and robustness of our proposed approach, we use two different NLP domain adaptation datasets, i.e., sentiment analysis corpus [30] and consumption intention corpus [4]. We will introduce the details of datasets in the followings.…”

“…Numerous natural language processing (NLP) tasks are domain-dependent, such as sentiment analysis [1], [2] and consumption intention mining [3], [4]. For such practical tasks, the performance of supervised learning models with insufficient training data cannot be satisfactory for applications, while manually annotating sufficient training data for all application domains is an almost impossible mission.…”