22 August 2014

SecondEGO - Dashboard/Status

Here is the latest tab on SecondEGO Dashboard. On this tab you can see some basic information about your chatbot. You can also disable/enable it. But the most important thing on this tab is light indicator.

A small light shines near the tab name (upper right). It indicates the status of your chatbot. The explanation of indicator is written with the same color, so it is easy understandable. Here is an example for blue light. Blue light occurs immediately after chatbot creating and shines until you insert embed code into your website.

If you have made any changes to your chatbot, they won't be visible to your visitors until you publish them. The orange color indicates this. But changes will be visible in your Dashboard/Test tab immediately , so you can test them and work on your chatbot improvement.  Here is an example for orange light:

After approval that all changes works fine, you can simply press Publish changes button. This will push all your changes to be available to visitors of your website. Since this moment, the green light shines near Status.

With this post we have finished with describing the main feature on SecondEGO Dashboard. 

SecondEGO - Dashboard/Analytics

After chatbot is implemented on your website, analytics is very handy tool to check what is going on. You won’t just see what questions was unanswered, you will also see what visitors are really searching for and what problems they have. In today’s world this is one of the most important information for website owners. This kind of information isn't available on Google Analytics, but only on customers support tools like SecondEGO.
Wouldn't be nice to know what items does website visitor miss? Or what kind of problem does he have with your product?
With these answers you can improve your website – add a new items, add solution for known problem, tell visitors how to use it…
And what do you think – where their next purchase will be done if they can find right answers fast and anytime on your website?
And if you see that chatbot can’t answer to frequently questions, you will spend just some minutes to improve/add that answer and its knowledge will helps to answer your visitors next time.

There are six tabs under Dashboard/Analytics:
- activity (real time and previous usage statistics),
- questions (all visitor’s questions),
- unrecognized questions (Questions where chatbot didn't have an answer. Maybe it looks strange naming to someone, because chatbot can use Site search in this case - if it was available.),
- answers,
- triggers (how many times individual triggers was used),
- other (knowledge modules usage).
Because questions, unrecognized answers and answers is sorted by frequency, your effort into adding patterns can quickly be visible. Checking analytics regularly is an important key to get a better results with chatbot on your website.

SecondEGO – Dashboard/Train/Site search

Site search is very powerful function. We will describe it's fundamentals today. You can use this function only, if your chatbot belongs to Business or Trial package.

What is Site search anyway?
It is intelligent website text search which uses two different but complementary approaches:
1. Advanced text search on website. It is usual text search with word forms and synonyms and is improved with advanced language technologies. Chatbot's answer is one (or more) link to webpage(s).
2. Crammer technology. It is Cross-Language "Question Answering” system based on interlingua (in this case, developers employ interlingua from Amebis). Sentences from input knowledge are translated into interlingua, split into fragments and added to the database. To answer, the question is translated into interlingua and split into fragments, then the database is searched for appropriate facts. The answers are then translated from interlingua into natural language; they can be short (just a specific phrase) or long (in a sentence).
It is obviously that Crammer technology has some limitations. At the moment, its success rate is about 50%. The reasons are that some grammatical structures aren't supported yet, analyzer can come across unknown words, disambiguation…
If visitor’s question don’t has an answer in patterns (and in predefined knowledge modules), than your chatbot answer with Site search.
Here are two examples (left with Advanced text search results, right with Crammer technology result):
SecondEGO - Site search (advanced text search and crammer)
Before your visitors can use Site search, your website needs to be analyzed and indexed. This might take several hours or even days – it depends on number of pages on your website and number of chatbots that waiting for the same thing on server.
You can track status of indexing next to where URL address has been entered.
SecondEGO - Site search status
Because indexing is repeated every two weeks, eventual changes on your website will be included automatically.
Indexation can take very long, so it is clear that you can use Site search only on your websites. You can enter http://www.wikipedia.org, but process won’t start.
This is checked with embed code which must be on your website. You can’t put embed code on Wikipedia (well there are some people who can :-)) and this is a good safety catch.

If your chatbot isn't ready for deployment yet and you want to test Site search function before deployment, you can comment embed code. The SecondEGO engine checks only an API key on webpage source code.
Let’s write this again to be clear. If you want to use Site search function, follow next steps:
1. Insert embed code (could be commented) into your website.
2. Click the “Publish changes” button under the “Dashboard/Status” tab.
3. Check status next to used URL address to see when it is finished.

This topic is just a short description about Site search and its background. There are many things behind this technologies, but describing them can be too complex and probably out of our context.
In a nuts, Site search is very powerful and interesting feature. It can drastically quicken implementation of chatbot, because all the knowledge on your existing website will be easily available to chatbot.

21 August 2014

GPS road navigation - Navfree

One reader asked me, if I know Navfree navigation. Of course I know it, but I haven't wrote about it until today. So I take it today with me and here are my first conclusions. It is another representative from OpenStreetMap camp.

+ Excellent user interface.
+ After entering address you get a bunch of context items (even with Street View).
+ Free, but with ads and some payable options (voices, safety cameras...).
+ Excellent map feedback tool.
- Pinch-in and Pinch-out gestures don't work. You have to use plus and minus button to zoom. Annoying.
- There is no route planner. It is very hard to prepare (or check) a route away from current or last GPS location.
- Not so many points of interests (POI), but it has Google search and this improves search for POI. A found a bug when I selected a route to one POI. An error "Cannot find a route" occurs.
- After the route was finished and program restarted, I found a rest of colored route on map.
- Catastrophic route planning. This can be a result of bad OpenStreetMap road categorization. After some disaster failures, I turn Navfree off. It tells everything.
- Navfree is stubborn, it insist on his route. If you turn away, it simply won't change a route a long time. In the meanwhile it will rotate a screen as mad.
- I didn't see any line assist and it can be, that Navfree don't has it.
- Bad ETA (estimated time of arrival).

After good first impression, the Navfree flunks on its main thing - navigation.
Rating: 4/10 (check whole ratings table)