We automatically create a sentiment sensitive thesaurus using both labeled and unlabeled data from multiple source domains to find the association between words that express similar sentiments in different domains.

Figure 4: Effect of source domain unlabeled data .

The amount of unlabeled data is held constant, so that any change in classification accu-

Figure 5: Effect of target domain unlabeled data .

positive or negative sentiment) given a small set of labeled data for the source domain, and unlabeled data for both source and target domains.

We use labeled data from multiple source domains and unlabeled data from source and target domains to represent the distribution of features.

Unlabeled data is cheaper to collect compared to labeled data and is often available in large quantities.

where y,-’ is the unobserved class label for the i-th instance in the unlabeled data .

By further considering the weight to ascribe to the unlabeled data vs. the labeled data (and the weight for the L2-norm regularization), we get the following regularized joint log likelihood to be maximized:

where the first term on the right-hand side is the log likelihood of the labeled data from both D1 and D2; the second is the log likelihood of the unlabeled parallel data U, multiplied by Al 2 O, a constant that controls the contribution of the unlabeled data ; and x12 2 0 is a regularization constant that penalizes model complexity or large feature weights.

Experiments on multiple data sets show that the proposed approach (1) outperforms the monolingual baselines, significantly improving the accuracy for both languages by 3.44%-8.l2%; (2) outperforms two standard approaches for leveraging unlabeled data ; and (3) produces (albeit smaller) performance gains when employing pseudo-parallel data from machine translation engines.

MaxEnt: This method learns a MaxEnt classifier for each language given the monolingual labeled data; the unlabeled data is not used.

SVM: This method learns an SVM classifier for each language given the monolingual labeled data; the unlabeled data is not used.

maximum entropy and SVM classifiers) as well as two alternative methods for leveraging unlabeled data (transductive SVMs (Joachims, 1999b) and co-training (Blum and Mitchell, 1998)).

To our knowledge, this is the first multilingual sentiment analysis study to focus on methods for simultaneously improving sentiment classification for a pair of languages based on unlabeled data rather than resource adaptation from one language to another.

Another line of related work is semi-supervised learning, which combines labeled and unlabeled data to improve the performance of the task of interest (Zhu and Goldberg, 2009).

We consider a semi-supervised setting for domain adaptation where only unlabeled data is available for the target domain.

For every pair, the semi-supervised methods use labeled data from the source domain and unlabeled data from both domains.

Also, it is important to point out that the SCL method uses auxiliary tasks to induce the shared feature representation, these tasks are constructed on the basis of unlabeled data .

In addition to the labeled data from the source domain, they also exploit small amounts of labeled data and/or unlabeled data from the target domain to estimate a more predictive model for the target domain.

In this paper we focus on a more challenging and arguably more realistic version of the domain-adaptation problem where only unlabeled data is available for the target domain.

(2006) use auxiliary tasks based on unlabeled data for both domains (called pivot features) and a dimensionality reduction technique to induce such shared representation.

Intuitively, maximizing the likelihood of unlabeled data is closely related to minimizing the reconstruction error, that is training a model to discover such mapping parameters u that z encodes all the necessary information to accurately reproduce :13“) from z for every training example :3“).

and unlabeled data for the source and target domain {m(l)}l€3UuTU, where SU and TU stand for the unlabeled datasets for the source and target domains, respectively.

However, given that, first, amount of unlabeled data |SU U TU| normally vastly exceeds the amount of labeled data |SL| and, second, the number of features for each example |a3(l)| is usually large, the label y will have only a minor effect on the mapping from the initial features a: to the latent representation z (i.e.

(2010), which introduces a high-level rule language, called NERL, to build the general and domain specific NER systems; and 2) semi-supervised learning, which aims to use the abundant unlabeled data to compensate for the lack of annotated data.

Semi-supervised learning exploits both labeled and unlabeled data .

It proves useful when labeled data is scarce and hard to construct while unlabeled data is abundant and easy to access.

Another representative of semi-supervised learning is learning a robust representation of the input from unlabeled data .

This is implemented by maximizing the empirical accuracy on the prior knowledge (labeled data) and the entropy of hash functions (estimated over labeled and unlabeled data ).

Let XL 2 {(X1,cl)...(xu,cu)} be the labeled data, c E {1...0}, X 6 RM, and XU = {xu+1...xN} the unlabeled data .

the labeled and unlabeled data , 26,; and XU.

Co-training puts features into disjoint subsets when learning from labeled and unlabeled data and tries to leverage this split for better performance.

For an unsupervised approach, which only needs unlabeled data , there is little cost to creating large training sets.

We also compare our method with the semi-supervised approaches, the semi-supervised approaches achieved very high accuracies by leveraging on large unlabeled data directly into the systems for joint learning and decoding, while in our method, we only explore the N-gram features to further improve supervised dependency parsing performance.

Some previous studies also found a log-linear relationship between unlabeled data (Suzuki and Isozaki, 2008; Suzuki et al., 2009; Bergsma et al., 2010; Pitler et al., 2010).

All of our selectional preference features described in this paper rely on probabilities derived from unlabeled data .