Interview with WhiteSmoke

Here is an interview I took with Hilla Ovil-Brenner, CEO of WhiteSmoke Inc.
Torsten:
Could you please tell us something about the origin of name "WhiteSmoke", how did it emerge?

Hilla Ovil-Brenner:
Well, as WhiteSmoke is a software aiding people to communicate in English, we looked for a concept that relates to communication. The Native American usage of smoke puffs to communicate alerts across vast distances and language groups fit just perfectly. This traditional practice reflects the way our software enables our global users to communicate clearly and effectively, overcoming any English language difficulties encountered by native speakers of different languages."
Torsten:
So the name Whitesmoke refers to the way people from different parts of the world exchange their thoughts with each other. Who founded the company and when?

Hilla Ovil-Brenner:
Hilla Ovil-Brenner, who was raised in South Africa, established WhiteSmoke in 2002 with husband Liran Brenner after her father complained about the poor quality of English in e-mails he received from foreign colleagues. Brenner, a computer engineer, brought with him five years experience at instant messaging pioneer ICQ. From humble beginnings in their apartment with a small investment and a lot of enthusiasm, WhiteSmoke is now a 30+ employee operation of international scope, and continues to grow monthly.

Torsten:
A lot of our users ask us how they can improve their writing skills. What do you think is the best method to learn how to compose effective and grammatically correct emails, memos, sales letters and other business documents?

Hilla Ovil-Brenner:
According to my experience, the best way to improve your writing skills is to practice writing as much as possible and get feedback on what you write. It’s true that you can always take English language or writing courses, and even programs like Microsoft Word can be of some help with their limited text correction features. Of course, I use WhiteSmoke when writing emails and business documents, and it is great for picking up hard-to-find grammar and spelling errors, and for suggesting words when I get stuck. I also work with templates for serious business documents -- having a well-written model when sitting down to write is an excellent tip for producing more professional texts. But ultimately, there is no one “best method” -- everyone needs to find their own, but I stand behind WhiteSmoke!"

Torsten:
I understand that there different versions of Whitesmoke for different purposes and markets. Do you have a primary target group?

Hilla Ovil-Brenner:
WhiteSmoke has a very broad “target group” – anybody writing in English! We designed it to cater to any level of English proficiency, but we also know that writers in different professional fields have specific needs. These are usually related to word choice and register rather than grammar, which is more universal. That’s why our R&D teams developed specially tailored versions for business, legal, medical, creative, and academic writing styles. Each version works with a different database and dictionary containing field-specific language, such as legal terms or business jargon. This is the source for context-specific spelling corrections and text enrichment suggestions. A general version is available for the home user, and a very popular executive package for more professional users includes all the other versions’ features.

Torsten:
Which version of WhiteSmoke would be best for users who speak English as second language?

WhiteSmoke:
The choice of version depends on the needs of the individual ESL learner. Of course, ESL learners do make different kinds of mistakes to many native speakers, but these are all covered by the grammar engine, which is common to all versions. The differences mainly lie in text enrichment. So, let’s say a person studies English for general use, then the general version could be enough. But if you want to focus on, for example, Business English, then your best option would be the business version, since its enrichment suggestions are totally relevant to business jargon – its algorithms are trained with business texts. This helps ESL learners with an interest in business to expand their vocabulary in the field. Likewise for the other versions – the ESL learner can use a version of WhiteSmoke closely tailored to his or her requirements.

Torsten:
Is it possible for a native speaker of English to distinguish between a text that has been written without Whitesmoke and one that was created with the software? I mean, how natural does a text sound that was composed with the help of Whitesmoke?

Hilla Ovil-Brenner:
Well, I think that a native speaker will not necessarily be able to distinguish whether a text was written with WhiteSmoke or not. Maybe if the reader is familiar with the writer’s level of English, like with English teachers and their students, it would be possible. The less proficient a writer is in English, the less natural the text will look. When a non-native English speaker writes, there are bound to be more mistakes on many text levels. Our software corrects errors of grammar, spelling, and punctuation at the level of the sentence, and really does a great job doing that. But a text is actually a very complex structure with many layers. There are things like idiomatic language, more complex sentence structures, and paragraph organization that no software is able to improve. This is why how natural the text sounds depends on the writer’s level, with or without the editing features WhiteSmoke currently offers."

Torsten:
On your website you mention that Whitesmoke uses patented natural language procession technology. How exactly is Whitesmoke's NLP technology different from that other text processing software programs?

Hilla Ovil-Brenner:
Well, I am always proud to say that the beauty of WhiteSmoke is that it uses state-of-the-art technology unique only to us! Other text editing software programs are based on a pre-dictated list of grammar and spelling rules. They simply compare the user’s text to these, and mark basic mistakes. But we’ve challenged this approach by creating smart artificial intelligence algorithms that perform both rule- and statistical-based analysis of real mistakes users actually make in real life. Any language expert knows that language can generate an infinite number of sentences with an infinite number of potential mistakes. No list can cover all the mistakes, so why not focus on what people actually write, and try to work with that? Because the algorithms are trained by scanning texts, the resulting text changes relate to how words relate to one another – the context, not just micro-level simple mistakes! This kind of outside-the-box thinking has also had some unexpected positive results, for instance, we managed to rank the most common mistakes and re-direct research efforts. We also saw that when we trained our patented algorithms on texts from different professional fields, such as business or law, we got different results for text enrichment. This made our numerous professional versions possible.

Torsten:
I've heard from some Whitesmoke users that if you launch the program from within MS Word, you can't edit your Word document since Whitesmoke works in plain text mode only. Is that true?

Hilla Ovil-Brenner:
Yes, that’s true. If a user writes a document in any word processor and then wants WhiteSmoke to work on the text, pressing F2 locks the original word processing software. This actually helps to prevent synchronization problems between the original document and the WhiteSmoke edit window, and is not related to using plain text mode. After the user saves the changes in the edit window and closes it, the original text gets automatically updated. The user can then continue writing another piece of text in the word processor and repeat the same process. This creates an organized workflow. According to our experience, a situation where users were able to edit text in two programs simultaneously would be very problematic technically, and simply confusing.

Torsten:
When I used the trial version of WhiteSmoke I noticed that the text enrichment feature often suggests words that don't necessarily fit into the given context. How is a user who speaks English as a second language supposed to know which of the 'enrichment' options they should select?

Hilla Ovil-Brenner:
Let me explain a little about how text enrichment works. Text Enrichment as WhiteSmoke implements it comprises two areas - "synonyms" and "adverbs and adjectives". Until the latest version of WhiteSmoke (released in the last month), alternative synonyms, adverbs, and adjectives were automatically offered to users based on WhiteSmoke's analysis of the text.
The new version of WhiteSmoke has introduced "Enrichment on Demand", which runs in conjunction with the standard text enrichment suggestions. For users, this means that synonyms can be selected manually for practically every word, giving them greater control over the feature, to use or not use according to their needs. The standard automatic enrichment feature has been revamped to compliment only highly targeted words with adverbs and adjectives. Our testing found that this approach greatly improved the experience of enrichment for native and non-native speakers alike.
Of course, making a selection between a number of synonyms, adverbs and adjectives may not be an easy choice for some English as a second language speakers. However, it is important that we do offer the capability, as it is a very useful feature for our large native-speaking customer base. English as a second language speakers who are using WhiteSmoke can also keep in mind that Text Enrichment is there as an aid, but it is not correcting mistakes, like the spelling and grammar engines. In cases where a user is unsure about the relevancy of a particular enrichment suggestion, he or she can simply ignore the suggestion.