There are many simple tasks that computers can't do well. Below are some examples of how txteagle users can help.
The global transcription market is expected to grow to 18 billion dollars by 2010. With currently over $12B in annual revenue, standard rates in the medical transcription industry are $100 for every 1,000 lines of text, or approximately 10 cents per 65 character line, with a 98% average accuracy rate. Using SMS concatenation, we have shown that 5 lines of audio text can be written down by hand and then copied into an SMS in less than 2 minutes. Paying proficient users $3/hour to do this work on their mobile phone drops the cost to 2 cents per line. Even collecting 100% overhead to cover payments to the operator and our other partners, this still results in a price reduction of over 60% from the today’s transcription rates, corresponding to an annual savings within the medical transcription industry of over 7 billion dollars.
There are over 100 distinct languages in Kenya alone. It is impossible for software companies to incorporate these languages into their interfaces because no translation service exists currently. We are generating a 'phrase book' of relevant words in every Kenyan language and will license this corpus to companies interested in localizing their software for the Kenyan market.
While the idea of citizen journalism is not new, it has yet to have significant penetration within the developing world due to low connectivity and literary rates. Additionally, the need for compensation is all the more critical to aspiring journalists in the developing world, especially due to the cost of connectivity to submit a news story. By leveraging our existing mobile payment system we enable anyone with a phone to submit a story for free and have compensation directly transmitted back to the handset. We envisage a huge demand for this service: there are tens of thousands of askari (guards) that sit out on the streets of Kenya all day – these are the guys that see the news first. txteagle empowers them to tell (and sell) their story to the rest of the country.
The major players in the multi-billion dollar internet search industry all have similar approaches to generating search results and as a result, all perform fairly similarly. By augmenting the machine learning search algorithms with human input, it may become possible to make dramatic improvements in relevancy for specific queries.
There are an increasing number of searches in tribal languages originating in remote villages. When a Kenyan outside of Kilifi village searches for “titanic bongo”, the results should be dramatically different from what is generated from today’s search engines. Txteagle enables search engines to evaluate the relevancy of their responses to these types of queries using the only people qualified to do so: the Kenyans themselves.
[Some of these search results are brought to you by Ahmed M., age 12, Tefera, Mozambique. He is using the money he has earned to help pay for his younger brother’s primary school education.]
Whether as a fee service or a method of differentiating one search engine from the rest, we will open our API to provide compelling details about the individuals earning money for completing each particular task. There will be a certain segment of search engine users who will choose to use a txteagle powered search services because of the desire to help provide an income source to those who need it most.
Major products get mentioned online thousands of times every day in many different languages. Gauging the sentiment of what is being written about a particular product is critical information for branding and marketing departments, yet extremely difficult to quantify in an automated way using traditional artificial intelligence techniques. Txteagle will ask users if a section of web-scraped text expresses a positive, negative, or neutral sentiment about a particular product or brand. When aggregated over time, this service will be able to provide detailed insight into the dynamics of fickle online public opinion and perception.
Conducting large demographic surveys in remote areas of the developing world is a challenging and costly endeavor. However in countries such as Kenya, where over 90% of the population lives in an area with GSM reception, it becomes possible to remotely administer surveys using text messaging. SMS surveys are the most efficient and economical method of collecting a wide variety of information including census, health, activity, and commerce data.
We'll soon be launching txteagle voice tasks in Uganda during off-peak hours when the network usage is significantly below capacity. Our first voice tasks will be rating radio commercials; these ratings will be incorproated into our collaborative filtering algorithms to provide subscribers with targeted advertisements. Additionally, the mobile web is growing dramatically in many parts of the developing world. Whether it is auditory information for local language transcription, image recognition tasks for video security camera monitoring, or simply more extensive text analysis, txteagle is currently working with operators to offer bandwidth rich services during times when their GPRS/EDGE network is underutilized.
Got other ideas for potential tasks? Let us know.
Client: Nokia
Target Subscriber: Coastal Kenya
Task: Software Localization
Question: Translate the phrase "Address Book" into Giriama.
-----
Client: New York General Hospital
Target Subscriber: Literate English Speaker
Task: Transcription (Audio/SMS Task)
Question: Transcribe the following audio clip from a New York hospital.
-----
Client: World Health Organization (WHO)
Target Subscriber: Coastal Kenya
Task: Malaria Survey
Question: Does your family regularly sleep under a bed net?
Reply: 1.Every night 2.Most nights (4-6 times / week) 3.Occasionally (2-3 times / week) 4.Rarely (1-5 times / month) 5.Never
-----
Client: Nokia Research Center
Target Subscriber: Luo Speaker
Task: Train a speech recognition engine: Listen and Repeat (Audio Task)
Text: Repeat the following Luo phrase and then press the * button.
-----
Client: Red Cross, Rwanda
Target Subscriber: Butare Resident
Task: Aid Efficacy Assessment
Question: There is a temporary hospital in the Biryogo district by the local market for residents injured during the Butare earthquake last night. Please inform others! Were you previously aware of this emergency facility?
Reply: 1.No, this is the first time I have heard about the hospital. 2.Yes, I already found out about the hospital though other means.
-----
Client: Kenyan Ministry of Health
Target Subscriber: Rural Nurses
Task: Assess Blood Supply Levels
Query: Please respond with the O-negative blood supply levels within your local district hospital bloodbank.