Long short-term memory (LSTM) networks have recently shown remarkable performance in several tasks that are dealing with natural language generation, such as image captioning or poetry composition. Yet, only few works have analyzed text generated by LSTMs in order to quantitatively evaluate to which extent such artificial texts resemble those generated by humans. We compared the statistical structure of LSTM-generated language to that of written natural language, and to those produced by Markov models of various orders. In particular, we characterized the statistical structure of language by assessing word-frequency statistics, long-range correlations, and entropy measures. Our main finding is that while both LSTM- and Markov-generated texts can exhibit features similar to real ones in their word-frequency statistics and entropy measures, LSTM-texts are shown to reproduce long-range correlations at scales comparable to those found in natural language. Moreover, for LSTM networks, a temperature-like parameter controlling the generation process shows an optimal value - for which the produced texts are closest to real language - consistent across different statistical features investigated.

Natural Language Statistical Features of LSTM-Generated Texts / LIPPI, MARCO. - In: IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS. - ISSN 2162-237X. - ELETTRONICO. - 30:(2019), pp. 8681285.3326-8681285.3337. [10.1109/TNNLS.2019.2890970]

Natural Language Statistical Features of LSTM-Generated Texts

LIPPI, MARCO
2019

Abstract

Long short-term memory (LSTM) networks have recently shown remarkable performance in several tasks that are dealing with natural language generation, such as image captioning or poetry composition. Yet, only few works have analyzed text generated by LSTMs in order to quantitatively evaluate to which extent such artificial texts resemble those generated by humans. We compared the statistical structure of LSTM-generated language to that of written natural language, and to those produced by Markov models of various orders. In particular, we characterized the statistical structure of language by assessing word-frequency statistics, long-range correlations, and entropy measures. Our main finding is that while both LSTM- and Markov-generated texts can exhibit features similar to real ones in their word-frequency statistics and entropy measures, LSTM-texts are shown to reproduce long-range correlations at scales comparable to those found in natural language. Moreover, for LSTM networks, a temperature-like parameter controlling the generation process shows an optimal value - for which the produced texts are closest to real language - consistent across different statistical features investigated.
2019
30
3326
3337
Goal 3: Good health and well-being
LIPPI, MARCO
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificatore per citare o creare un link a questa risorsa: https://hdl.handle.net/2158/1356466
Citazioni
  • ???jsp.display-item.citation.pmc??? 4
  • Scopus 69
  • ???jsp.display-item.citation.isi??? 55
social impact