Greedy Intelligence’s 1Checker Is A Proofreading Platform For Non-Native English Speakers

Developed by two Cambridge PhD students, 1Checker is a proofreading platform that seeks to be a better alternative to Microsoft Word’s spellchecker and Grammarly for non-native English speakers. 1Checker and its sibling 1Course, an online learning management platform for language schools, identifies errors commonly made by people who learned English as a second language and makes contextually appropriate suggestions so users can avoid embarrassing errors.

Both software platforms were developed by Greedy Intelligence, a London-based startup that received seed funding from Cambridge Enterprise, the university’s seed fund, and is currently closing its Series A round.

Co-founder Yichi Zhang began working on 1Checker’s natural language processing technology in 2007 while studying for a PhD in software engineering at Cambridge. Zhang, a native Chinese speaker, had a difficult time finding proofreaders to help him check academic papers he wanted to submit for publication. He also discovered that Microsoft Word’s built-in spelling and grammar checking tools frequently did not catch common errors made by non-native English speakers.

For example, Zhang says, someone who learned English as a second language might write “red big bag” instead of “big red bag.” Word’s spellchecker would ignore the confusing sequence of adjectives because it is a mistake most native English speakers don’t make. Other common errors include mixing up homophones like “affect” and “effect.”

1Checker screenshot

Zhang developed 1Checker with Greedy Intelligence CTO Lin Sun and other Cambridge graduate students from China who wanted to use the software on their own writing. The team tried different algorithms before having a “major breakthrough” in 2009, when they finally achieved the right balance between coverage and accuracy, the two things proofreading software has to balance.

“We maintained a certain level of accuracy while we expanded coverage to make sure that the checker could understand errors by non-native speakers,” says Zhang. “We used machine learning and other artificial intelligence algorithms to make the computer not only understand grammar rules, but to also understand the semantics and context of the whole paper. That’s why the system can spot contextual errors.”

In 2010, Greedy Intelligence won the £5,000 Cambridge University Entrepreneurs’ (CUE) challenge and received funding from Cambridge Enterprise, the university’s official seed funding, marking its transformation from an extracurricular project into a startup. Earlier this year, the company launched 1Course, its first paid product. 1Course is targeted toward language schools that offer prep courses for the TOEFL, GRE, GMAT and other standardized tests. Many of these institutions use scoring software like ETS’ e-rater or Vantage IntelliMetric to check student essays for errors.

1Course seeks to be better at evaluating how clearly non-native English speakers have expressed their thesis, supporting material and conclusion.

“You have to understand the organization, development and grammar, as well as the complexity of the vocabulary and sentence structures,” says Zhang.

Improving Accuracy

One of the reasons Greedy Intelligence offers 1Checker for free is so the software can glean more data and refine its algorithm. Both 1Checker and 1Course become more accurate as more people use it, but Zhang says Greedy Intelligence is also mindful of user privacy. It learns from user preference (for example, which vocabulary suggestions people pick the most often), but it doesn’t keep their writing.

“It logs users’ choices rather than their sentences,” says Zhang. “We definitely don’t know the user’s tasks or their private information and, to be honest, we’re not that interested.”

Most of Greedy Intelligence’s users are currently located in China or countries such as Malaysia and Singapore where English is frequently spoken as a second language. The company wants to begin marketing its software in other languages and will re-invest earnings from 1Course in research and development. The technology used by Greedy Intelligence is not language-related, so its basic algorithm and whole mechanism can be adapted to other languages fairly quickly, says Zhang. 1Checker currently has a total of 700,000 users.

“That is not a lot compared to other free software available in China, but it’s a very niche market. We are targeting non-native speakers, most of them potential overseas students, including some academic speakers that want to get published in international conference journals, as well as people working for international companies,” says Zhang.

The company plans to add features such as “phase level enrichment,” which means that 1Checker will help users rephrase sentences that are confusing or too colloquial. Support will also be added for mathematical papers and legal terms.

“We want to offer features that no language tools or grammar checkers currently do. Some can, but its human-coded grammar rules, not results based on computer intelligence,” says Zhang.

Gaining Traction

Both 1Course and 1Checker are based on the same technology. Zhang identifies 1Checker’s main competitors as Microsoft Word’s spelling and grammar checking tools, as well as online proofreading platforms like Grammarly, which also offers a plagiarism checking feature. Both of Greedy Intelligence’s products seek to differentiate by accurately honing in on errors made by non-native speakers, as well as offering language enrichment tools.

“1Checker has much, much wider coverage, especially with errors made by non-native speakers, simply because we use non-native models. Our software is designed especially for non-native speakers. We have certain corpora, the linguistics term for language data. We collect a lot of language data written by non-native speakers, as well as some corrected by non-native speakers,” says Zhang. “Our development team understands the pain of non-native speakers, too.”

1Checker and 1Course’s enrichment tools are designed to help its users quickly improve their English writing skills. For instance, if you use a certain word too many times, 1Checker will automatically suggest new or more advanced vocabulary. Most importantly for non-native English speakers, the suggested words are contextually appropriate. The next update of 1Checker will add grammar enrichment so the software can suggest suitable alternatives for overused phrases.

“If you use StyleWriter, Grammarly or other proofreading tools, they will only tell you if you are using simple sentences too much,” says Zhang. “What a non-native speaker needs are very detailed suggestions. If I am writing a paper or an email, I probably don’t have hours to proofread or rewrite. The only thing I want to do is spend five or 10 minutes making my writing as professional as possible.”