Computer that can tell the write sex just by reading

A COMPUTER programme has been developed which can distinguish our sex simply by looking at the way we write.

But rather than looking at words for their macho posturing or feminine sense, the Israeli-developed programme has confirmed what many scholars already thought - women focus on people while men prefer to concentrate on things.

To some - including an academic journal which refused to publish the research - the assumption may be sexist but to the scientists behind the project, the microchips processing the data were beyond reproach.

Hide Ad
Hide Ad

Moshe Koppel, a professor at Bar Ilan University in Israel, says his computer algorithm can detect to an accuracy of 80 per cent the sex of an author.

Prof Koppel’s team fed 604 texts taken from the British National Corpus, a collection of 4,124 documents assembled by academics to study language use, into the programme.

From a biography of Kylie Minogue to novels such as Talking it Over by Julian Barnes, the texts are divided equally between male and female authors.

The research found that the single biggest difference is that women are more likely to use personal pronouns than men.

While a woman will use "I", "you" and "she", men are more prone to terms that include "a", "the" and "these", as well as numbers and quantifiers such as "more" or "some".

Prof Koppel’s team noted that key words used to describe cars in Top Gear Magazine or flowers in Homes and Gardens magazine were not as important in determining sex as smaller, component parts of text which, at an unconscious level, revealed gender-specific patterns.

However, the announcement that computers might distinguish our sex has not gone been welcomed in some academic quarters.

Prof Koppel said that when his research group first submitted one of two papers to be published to the publishing panel of the prestigious National Academy of Sciences in the United States, the referees rejected it "on ideological grounds". "They said, ‘What do you mean? You’re trying to make some claim about men and women being different, and we don’t know if that’s true. That’s just the kind of thing that people are saying in order to oppress women’," he told The Boston Globe.

Hide Ad
Hide Ad

Prof Koppel, whose team included a female academic, retorted that he was "just reporting the numbers".

Using the British texts, Prof Koppel said the process was almost flawless. After the artificial intelligence had sampled its core 604 texts, the other tomes were fed into the algorithm to determine the elements in writing unique to either men or women. "The more frequently a word got used in one set, the more weight it got," Prof Koppel explained. "If the word ‘you’ got used in the female set very often and not in the male set, you give it a stronger female weighting." When the process was complete, the scientists found a set of 50 features, which the algorithm - when applied to text in the same way that a word-count process is on a editing programme - could use to tell an author’s gender.

In an attempt to silence critics, Prof Koppel’s team analysed scientific papers to see if the system would still work.

"If the argument that content belies the author held true then the computer would find it more difficult to distinguish gender in fairly flat, academic prose, but the computer won. It blew my mind," said Prof Koppel.

However, the application of the programme may yet hit a snag. Critics claim that experiments in gender-prediction can’t discover inalienable differences simply based on a male or female model. Janet Bing, a linguist with the Old Dominion University in Norfolk, Virginia, said that lesbian, gay, bisexual and transgendered people did not fit adequately into a simple social definition of male or female gender. She added: "This whole rush to categorisation usually works against women," she told the Globe.

"You find what you’re looking for and that leads to this sneaking suspicion that it’s all hardwired instead of cultural."

However, a fellow linguist and author, Deborah Tannen, disagreed, saying the research was "not too surprising". She said: "What are personal pronouns? They’re talking about people and we know that women write more about people."

Ms Tannen, who has written about the differences in men and women’s speech in conversations, said that women typically write in a more "involved" style, trying to encourage a more intimate connection with the reader and leading to greater use of pronouns.

Related topics: