Hello,
we are a software company with the goal to make computer better in understanding humans. For this working with lingubots is part of what we want to start with. (We are doing business for over 10 years…)
I really honor the work Bruce has done and giving out manuals and source is a great help.
Nonetheless it is very hard to get into the script. I miss a lot of samples to examine and best practices…
We are now talking with Bavarian Government to create a lingubot that helps people to find the right contact within the administration and to get some basics facts. People should ask “Where do I have to register my car and what do I have to take with me if I live in village xy” or “I need financial support. I have a family with 6 children. What can you do for me”.
I believe a chatbot that mixes answers to certain topics (written in chatscript) with answers that come from a web service (what are the office hours for car registration in munich) would help.
I wonder if I can put all information about office in facts or if I should write a script.
AND: This should be all in German.
So here are my questions:
Has anyone a German DICT?
Would you put office information in facts or in web service? Any idea how to join chatscript with web services?
Where are some sample top files? Or is it possible to have a consultant guide me?
Best regards

Bernhard

Most likely you would want a mixture of facts, script, and “web services” (by which I presume you mean search but I’m not sure). The problem with search is getting back more data than you want. The advantage is you didn’t have to prepare in advance for the question. The advantage of facts is compact data storage and creation. The advantage of script is natural human-like conversation.

There is a german wordnet dict, but it is not free. I have no idea what it costs to license it. One can probably live without it as follows:  One use of the dictionary is to support automatic spelling correction. Any online list of the 20K or so most common german words would suffice to create that ability. Another use of the dictionary is to provide an enumeration of relevant words for predefined chatscript concepts.  Like ~food can list a bunch of words, and also list a particular wordnet meaning and thus get all words below it in the wordnet ontology automatically. But for your application, you are not likely to need that. The ontologies needed for your application can be replicated in german from the existing english. Eg, you will probably want to replicate the verb concept lists, but wordnet’s verb ontology is basically useless and so chatscript doesn’t care about it. Adverb and Adjective ontologies same thing.  The noun ontologies of wordnet are useful for many things, but for the kinds of questions you will be fielding, you will probably be building your own german noun concepts anyway.

To build a chatscript server that is fast that can involve web services, the likely architecture is:
1. you have your own server to take the incoming user request
2. you pass that over to a chatscript server to understand what the user wants
3. chatscript sends back to your server a combination of “chat” and “out-of-band” commands like “search on this”.
4. your server performs the command and sends the results plus chat back to the user.

I am available for consulting work at various times. Contact me privately by email for further details.