Ting Liu

PENS: a machine-aided english writing system for Chinese users

Writing English is a big barrier for most Chinese users. To build a computer-aided system that helps Chinese users not only on spelling checking and grammar checking but also on writing in the way of native-English is a challenging task. Although machine translation is widely used for this purpose, how to find an efficient way in which human collaborates with computers remains an open issue. In this paper, based on the comprehensive study of Chinese users requirements, we propose an approach to machine aided English writing system, which consists of two components: 1) a statistical approach to word spelling help, and 2) an information retrieval based approach to intelligent recommendation by providing suggestive example sentences. Both components work together in a unified way, and highly improve the productivity of English writing. We also developed a pilot system, namely PENS (Perfect ENglish System). Preliminary experiments show very promising results.

Dependency Parsing Based on Dynamic Local Optimization

This paper presents a deterministic parsing algorithm for projective dependency grammar. In a bottom-up way the algorithm finds the local optimum dynamically. A constraint procedure is made to use more structure information. The algorithm parses sentences in linear time and labeling is integrated with the parsing. This parser achieves 63.29% labeled attachment score on the average in CoNLL-X Shared Task.

Topic tracking based on keywords dependency profile

Topic tracking is an important task of Topic Detection and Tracking (TDT). Its purpose is to detect stories, from a stream of news, related to known topics. Each topic is known by its association with several sample stories that discuss it. In this paper, we propose a new method to build the keywords dependency profile (KDP) of each story and track topic basing on similarity between the profiles of topic and story. In this method, keywords of a story are selected by document summarization technology. The KDP is built by keywords co-occurrence frequency in the same sentences of the story. We demonstrate this profile can describe the core events in a story accurately. Experiments on the mandarin resource of TDT4 and TDT5 show topic tracking system basing on KDP improves the performance by 13.25% on training dataset and 7.49% on testing dataset comparing to baseline.

Chinese Topic Link Detection Based on Semantic Domain Language Model: Chinese Topic Link Detection Based on Semantic Domain Language Model

Topic link detection is a foundational research in the field of topic detection and tracking, which detects whether two random stories talk about the same topic. This paper proposes a method of applying semantic domain language model to link detection, based on the structure relation among contents and the semantic distribution in a story, and also verifies the influence of the strategy incorporating dependency parsing into semantic description. Evaluation on Chinese Corpus of TDT4 show that the semantic domain language model substantially improved the performance of current detection system, whose minimum DET cost is reduced by about 3 percent.

Word sense language model for information retrieval

This paper proposes a word sense language model based method for information retrieval. This method, differing from most of traditional ones, combines word senses defined in a thesaurus with a classic statistical model. The word sense language model regards the word sense as a form of linguistic knowledge, which is helpful in handling mismatch caused by synonym and data sparseness due to data limit. Experimental results based on TREC-Mandarin corpus show that this method gains 12.5% improvement on MAP over traditional tf-idf retrieval method but 5.82% decrease on MAP compared to a classic language model. A combination result of this method and the language model yields 8.92% and 7.93% increases over either respectively. We present analysis and discussions on the not-so-exciting results and conclude that a higher performance of word sense language model will owe to high accurate of word sense labeling. We believe that linguistic knowledge such as word sense of a thesaurus will help IR improve ultimately in many ways.

Web mining for lexical context-specific paraphrasing

In most applications of paraphrasing, contextual information should be considered since a word may have different paraphrases in different contexts. This paper presents a method that automatically acquires lexical context-specific paraphrases from the web. The method includes two main stages, candidate paraphrase extraction and paraphrase validation. Evaluations were conducted on a news title corpus whereby the context-specific paraphrasing method was compared with the Chinese synonymous thesaurus. Results show that the precision of our method is above 60% and the recall is above 55%, which outperforms the thesaurus significantly.

Combining sentence length with location information to align monolingual parallel texts

Abundant Chinese paraphrasing resource on Internet can be attained from different Chinese translations of one foreign masterpiece. Paraphrases corpus is the corpus that includes sentence pairs to convey the same information. The irregular characteristics of the real monolingual parallel texts, especially without the strictly aligned paragraph boundaries between two translations, bring a challenge to alignment technology. The traditional alignment methods on bilingual texts have some difficulties in competency for doing this. A new method for aligning real monolingual parallel texts using sentence pairs length and location information is described in this paper. The model was motivated by the observation that the location of a sentence pair with certain length is distributed in the whole text similarly. And presently, a paraphrases corpus with about fifty thousand sentence pairs is constructed.

A Temporal Topic Model for New Event Detection

As an important task in Topic Detection and Tracking (TDT), New Event Detection (NED) aims to monitor the stream of news stories and detect new events reported by the first story of a topic. This paper proposes a Temporal Topic Model (TTM) to describe a topic as a series of events corresponding to different time. NED, based on TTM, firstly identifies whether a story includes the same time expressions with old topics, and then it verifies whether the story and the topics include relevant events corresponding to the expressions. Thus, a story will be determined as a new event, if it includes much few simultaneous relevant events of old topics. Additionally, this paper analyzes the distribution of time expressions to identify both seminal and novel events, by which NED modifies the probabilities of stories to be new events based on whether they include seminal or novel events of old topics. We compare our methods with some existing NED systems on TDT4 corpus, which demonstrate our methods substantially improve the efficiency and accuracy of NED.