Finding parts in very large corpora

Berland, Matthew; Charniak, Eugene

doi:10.3115/1034678.1034697

Cited by 276 publications

(196 citation statements)

References 10 publications

Supporting

Mentioning

192

Contrasting

Unclassified

Order By: Relevance

“…Patterns (symmetric or not) were found useful in a variety of NLP tasks, including identification of word relations such as hyponymy (Hearst, 1992), meronymy (Berland and Charniak, 1999) and antonymy (Lin et al, 2003). Patterns have also been applied to tackle sentence level tasks such as identification of sarcasm , sentiment analysis and authorship attribution (Schwartz et al, 2013).…”

Section: Related Workmentioning

confidence: 99%

“…In previous work patterns were used to represent a variety of semantic relations, including hyponymy (Hearst, 1992), meronymy (Berland and Charniak, 1999) and antonymy (Lin et al, 2003). Here, in order to capture similarity between words, we use Symmetric patterns (SPs), such as "X and Y" and "X as well as Y", where each of the words in the pair can take either the X or the Y position.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Symmetric Pattern Based Word Embeddings for Improved Word Similarity Prediction

Schwartz¹,

Reichart²,

Rappoport³

2015

Proceedings of the Nineteenth Conference on Computational Natural Language Learning

100

View full text Add to dashboard Cite

We present a novel word level vector representation based on symmetric patterns (SPs). For this aim we automatically acquire SPs (e.g., "X and Y") from a large corpus of plain text, and generate vectors where each coordinate represents the cooccurrence in SPs of the represented word with another word of the vocabulary. Our representation has three advantages over existing alternatives: First, being based on symmetric word relationships, it is highly suitable for word similarity prediction. Particularly, on the SimLex999 word similarity dataset, our model achieves a Spearman's ρ score of 0.517, compared to 0.462 of the state-of-the-art word2vec model. Interestingly, our model performs exceptionally well on verbs, outperforming stateof-the-art baselines by 20.2-41.5%. Second, pattern features can be adapted to the needs of a target NLP application. For example, we show that we can easily control whether the embeddings derived from SPs deem antonym pairs (e.g. (big,small)) as similar or dissimilar, an important distinction for tasks such as word classification and sentiment analysis. Finally, we show that a simple combination of the word similarity scores generated by our method and by word2vec results in a superior predictive power over that of each individual model, scoring as high as 0.563 in Spearman's ρ on SimLex999. This emphasizes the differences between the signals captured by each of the models.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Symmetric Pattern Based Word Embeddings for Improved Word Similarity Prediction

Schwartz¹,

Reichart²,

Rappoport³

2015

Proceedings of the Nineteenth Conference on Computational Natural Language Learning

100

View full text Add to dashboard Cite

show abstract

“…For example, extracting hyponyms [10], synonyms [13], meronyms [5] are specific instances of this general problem of relation extraction. Manually created or automatically extracted lexico-syntactic patterns have been successfully used to identify various relations between words [17,18].…”

Section: Related Workmentioning

confidence: 99%

Automatically Extracting Personal Name Aliases from the Web

Bollegala

Honma

Matsuo

et al. 2008

Advances in Natural Language Processing

View full text Add to dashboard Cite

Abstract. Extracting aliases of an entity is important for various tasks such as identification of relations among entities, web search and entity disambiguation. To extract relations among entities properly, one must first identify those entities. We propose a novel approach to find aliases of a given name using automatically extracted lexical patterns. We exploit a set of known names and their aliases as training data and extract lexical patterns that convey information related to aliases of names from text snippets returned by a web search engine. The patterns are then used to find candidate aliases of a given name. We use anchor texts to design a word co-occurrence model and use it to define various ranking scores to measure the association between a name and a candidate alias. The ranking scores are integrated with page-count-based association measures using support vector machines to leverage a robust alias detection method. The proposed method outperforms numerous baselines and previous work on alias extraction on a dataset of personal names, achieving a statistically significant mean reciprocal rank of 0.6718. Experiments carried out using a dataset of location names and Japanese personal names suggest the possibility of extending the proposed method to extract aliases for different types of named entities and for other languages. Moreover, the aliases extracted using the proposed method improve recall by 20% in a relation-detection task.

show abstract

“…-Natural language processing. We can utilize known techniques (e.g., [4]) to parse each sentence of a text or its summary obtained by means of text summarization [44]. For each resulting triple subject, predicate, object , we try to match it with the patterns s, p, o in the domain ontologies, where p is a property of the concept s and has a value typed of o.…”

Section: Associationmentioning

confidence: 99%

A layered framework supporting personal information integration and application design for the semantic desktop

Cruz

Xiao

2008

The VLDB Journal

View full text Add to dashboard Cite

Abstract. With the development of inexpensive storage devices, space usage is no longer a bottleneck for computer users. However, the increasingly large amount of personal information poses a critical problem to those users: traditional file organization in hierarchical directories may not be suited to the effective management of personal information because it ignores the semantic associations therein and bears no connection with the applications that users will run. To address such limitations, we present our vision of a semantic desktop, which relies on the use of ontologies to annotate and organize data and on the concept of personal information application (PIA), which is associated with a user's task. The PIA designer is the tool that is provided for building a variety of PIAs consisting of views (e.g., text, list, table, graph), which are spatially arranged and display interrelated fragments of the overall personal information. The semantic organization of the data follows a layered architecture that models separately the personal information, the domain data, and the application data. The network of concepts that ensues from extensive annotation and explicit associations lends itself well to rich browsing capabilities and to the formulation of expressive databaselike queries. These queries are also the basis for the interaction among views of the PIAs in the same desktop or in networked desktops. In the latter case, the concept of desktop service provides for a semantic platform for the integration of information across different desktops and the web. In this paper, we present in detail the semantic organization of the information, the overall system architecture and implementation aspects, queries and their processing, PIAs and the PIA designer, including usability studies on the designer, and the concepts of semantic navigation in a desktop and of interoperation in a network of desktops.

show abstract

Finding parts in very large corpora

Cited by 276 publications

References 10 publications

Symmetric Pattern Based Word Embeddings for Improved Word Similarity Prediction

Symmetric Pattern Based Word Embeddings for Improved Word Similarity Prediction

Automatically Extracting Personal Name Aliases from the Web

A layered framework supporting personal information integration and application design for the semantic desktop

Contact Info

Product

Resources

About