Data Mining Practical Machine Learning Tools and Techniques

Continuous word representations, trained on large unlabeled corpora are useful for many natural language processing tasks. Popular models to learn such representations  ignore the morphology of words, by assigning a distinct vector to each word. This is a limitation, especially for languages with large vocabularies and many rare words. In this paper, we propose a new approach based on the skipgram model, where each word is represented as a bag of character n-grams. A vector representation is associated to each character n-gram, words being represented as the sum of these representations. Our method is fast, allowing to train models on large corpora quickly and allows to compute word representations for words that did not appear in the training data. We evaluate our word representations on nine different languages, both on word similarity and analogy tasks. By comparing to recently proposed morphological word representations, we show that our vectors achieve state-of-the-art performance on these tasks.

/pdf/enriching-word-vectors-with-subword-information-22sqq6w40j.pdf

Enriching Word Vectors with Subword Information

Semantic word spaces have been very useful but cannot express the meaning of longer phrases in a principled way. Further progress towards understanding compositionality in tasks such as sentiment detection requires richer supervised training and evaluation resources and more powerful models of composition. To remedy this, we introduce a Sentiment Treebank. It includes fine grained sentiment labels for 215,154 phrases in the parse trees of 11,855 sentences and presents new challenges for sentiment compositionality. To address them, we introduce the Recursive Neural Tensor Network. When trained on the new treebank, this model outperforms all previous methods on several metrics. It pushes the state of the art in single sentence positive/negative classification from 80% up to 85.4%. The accuracy of predicting fine-grained sentiment labels for all phrases reaches 80.7%, an improvement of 9.7% over bag of features baselines. Lastly, it is the only model that can accurately capture the effects of negation and its scope at various tree levels for both positive and negative phrases.

/pdf/recursive-deep-models-for-semantic-compositionality-over-a-35h7g7u52l.pdf

Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank

Continuous word representations, trained on large unlabeled corpora are useful for many natural language processing tasks. Popular models that learn such representations ignore the morphology of words, by assigning a distinct vector to each word. This is a limitation, especially for languages with large vocabularies and many rare words. In this paper, we propose a new approach based on the skipgram model, where each word is represented as a bag of character $n$-grams. A vector representation is associated to each character $n$-gram; words being represented as the sum of these representations. Our method is fast, allowing to train models on large corpora quickly and allows us to compute word representations for words that did not appear in the training data. We evaluate our word representations on nine different languages, both on word similarity and analogy tasks. By comparing to recently proposed morphological word representations, we show that our vectors achieve state-of-the-art performance on these tasks.

We analyze skip-gram with negative-sampling (SGNS), a word embedding method introduced by Mikolov et al., and show that it is implicitly factorizing a word-context matrix, whose cells are the pointwise mutual information (PMI) of the respective word and context pairs, shifted by a global constant. We find that another embedding method, NCE, is implicitly factorizing a similar matrix, where each cell is the (shifted) log conditional probability of a word given its context. We show that using a sparse Shifted Positive PMI word-context matrix to represent words improves results on two word similarity tasks and one of two analogy tasks. When dense low-dimensional vectors are preferred, exact factorization with SVD can achieve solutions that are at least as good as SGNS's solutions for word similarity tasks. On analogy questions SGNS remains superior to SVD. We conjecture that this stems from the weighted nature of SGNS's factorization.

/pdf/neural-word-embedding-as-implicit-matrix-factorization-ltwfy3k99t.pdf

Neural Word Embedding as Implicit Matrix Factorization

Research into corpus-based semantics has focused on the development of ad hoc models that treat single tasks, or sets of closely related tasks, as unrelated challenges to be tackled by extracting different kinds of distributional information from the corpus. As an alternative to this "one task, one model" approach, the Distributional Memory framework extracts distributional information once and for all from the corpus, in the form of a set of weighted word-link-word tuples arranged into a third-order tensor. Different matrices are then generated from the tensor, and their rows and columns constitute natural spaces to deal with different semantic problems. In this way, the same distributional information can be shared across tasks such as modeling word similarity judgments, discovering synonyms, concept categorization, predicting selectional preferences of verbs, solving analogy problems, classifying relations between word pairs, harvesting qualia structures with patterns or example pairs, predicting the typical properties of concepts, and classifying verbs into alternation classes. Extensive empirical testing in all these domains shows that a Distributional Memory implementation performs competitively against task-specific algorithms recently reported in the literature for the same tasks, and against our implementations of several state-of-the-art methods. The Distributional Memory approach is thus shown to be tenable despite the constraints imposed by its multi-purpose nature.

/pdf/distributional-memory-a-general-framework-for-corpus-based-3qo0shtoyu.pdf

Distributional memory: A general framework for corpus-based semantics

We introduce BLESS, a data set specifically designed for the evaluation of distributional semantic models BLESS contains a set of tuples instantiating different, explicitly typed semantic relations, plus a number of controlled random tuples It is thus possible to assess the ability of a model to detect truly related word pairs, as well as to perform in-depth analyses of the types of semantic relations that a model favors We discuss the motivations for BLESS, describe its construction and structure, and present examples of its usage in the evaluation of distributional semantic models

/pdf/how-we-blessed-distributional-semantic-evaluation-2p94x8bmwg.pdf

How we BLESSed distributional semantic evaluation

Distributional semantics is a usage-based model of meaning, based on the assumption that the statistical distribution of linguistic items in context plays a key role in characterizing their semantic behavior. Distributional models build semantic representations by extracting co-occurrences from corpora and have become a mainstream research paradigm in computational linguistics. In this review, I present the state of the art in distributional semantics, focusing on its assets and limits as a model of meaning and as a method for semantic analysis.

Distributional Models of Word Meaning

The hypothesis that word co-occurrence statistics extracted from text corpora can provide a basis for semantic representations has been gaining growing attention both in computational linguistics and in cognitive science The terms distributional, context-theoretic, corpusbased or statistical can all be used (almost interchangeably) to qualify a rich family of approaches to semantics that share a “usage-based” perspective on meaning, and assume that the statistical distribution of words in context plays a key role in characterizing their semantic behavior Besides this common core, many differences exist depending on the specific mathematical and computational techniques, the type of semantic properties associated with text distributions, the definition of the linguistic context used to determine the combinatorial spaces of lexical items, etc Yet, at a closer look, we may discover that the commonalities are more than we could expect prima facie, and that a general model of meaning can indeed be discerned behind the differences, a model that formulates specific hypotheses on the format of semantic representations, and on the way they are built and processed by the human mind Methods for computational analysis of word distributional properties have been developed both in computational linguistics and in psychology Because of the different aims of each field, these lines of research have typically proceeded totally in a parallel fashion, often ignoring each other The drawbacks of this situation are clear: many

Distributional semantics in linguistic and cognitive research

The project LE-SIMPLE is an innovative attempt of building harmonized syntactic-semantic lexicons for twelve European languages, aimed at use in different Human Language Technology applications. SIMPLE provides a general design model for the encoding of a large amount of semantic information, spanning from ontological typing, to argument structure and terminology. SIMPLE thus provides a general framework for resource development, where state-of-the-art results in lexical semantics are coupled with the needs of Language Engineering applications accessing semantic information.

Alessandro Lenci

Papers

Distributional memory: A general framework for corpus-based semantics

How we BLESSed distributional semantic evaluation

Distributional Models of Word Meaning

Distributional semantics in linguistic and cognitive research

Simple: a general framework for the development of multilingual lexicons