Hi Jo,
I revised my proposal by adding the topic-based disambiguation method.
Actually, I found a mistake in my previous discussion on the topic-based
model. The topic-based model jointly models *the context
compatibility(which I missed)* and topic coherence. Thus, it can perform
well for both topic related and non-related entities. It is also a
generative model.
Thanks.
Regards,
Wei
On Mon, Apr 29, 2013 at 10:03 PM, Wang Wei <[email protected]> wrote:
> Hi Jo,
>
> I will think about the disambiguation part and revise my proposal.
> Thanks.
>
> Best Regards,
> Wei Wang
>
>
> On Mon, Apr 29, 2013 at 7:57 PM, Wang Wei <[email protected]> wrote:
>
>> Hi Jo,
>>
>> I read the proposal of the topic extraction work from last year's GSoC.
>> Yes, the topic-based disambiguation method is based on the LDA model. But
>> their objectives are different: topic extraction is to assign topic
>> categories for a document, while topic-based disambiguation is to
>> disambiguate entities based on the document's topic. For example, if a
>> document's topic is about 'mobile phones', then word 'Apple' would likely
>> be assigned as Apple Inc. . But, as I mentioned in my proposal, for topic
>> related entities, they can be disambiguated correctly; for other
>> entities, they are not guaranteed to be disambiguated correctly by
>> topic-based disambiguation method.
>>
>> In addition, for the generative model(the default disambiguation model),
>> it has a strong assumption: p(c|e)=p_e(t_1) p_e(t_2)...p_e(t_n), i.e., the
>> terms are independent given the entity e. Some improvements may be achieved
>> if this assumption is removed.
>>
>> Thanks.
>>
>> Best Regards,
>> Wei Wang
>>
>>
>> On Mon, Apr 22, 2013 at 2:30 AM, Joachim Daiber <[email protected]
>> > wrote:
>>
>>> Hey,
>>>
>>> you can have a look at Hector's github repository from last GSoC (this
>>> is not merged into the main branch yet):
>>>
>>> https://github.com/hunterhector/dbpedia-spotlight
>>>
>>> I think this was the paper he implemented:
>>>
>>> Han, X., 2011. Collective Entity Linking in Web Text : A Graph-Based
>>> Method. In *Proceedings of the 34th international ACM SIGIR conference
>>> on Research and development in Information*. pp. 765-774.
>>>
>>> Best,
>>> Jo
>>>
>>>
>>>
>>>
>>>
>>> On Sun, Apr 21, 2013 at 6:22 PM, Wang Wei <[email protected]> wrote:
>>>
>>>> Hi Jo,
>>>>
>>>> I am trying to learning something about the idea for "efficient graph
>>>> based disambiguation". However, the introduction is very short. Do you have
>>>> any further materials for the disambiguation methods used by db-pedia? In
>>>> the disambiguation code directory[1], which one is graph based?
>>>>
>>>> Thanks a lot.
>>>>
>>>>
>>>> Best Regards,
>>>> Wei Wang
>>>>
>>>>
>>>>
>>>
>>
>
------------------------------------------------------------------------------
Introducing AppDynamics Lite, a free troubleshooting tool for Java/.NET
Get 100% visibility into your production application - at no cost.
Code-level diagnostics for performance bottlenecks with <2% overhead
Download for free and get started troubleshooting in minutes.
http://p.sf.net/sfu/appdyn_d2d_ap1
_______________________________________________
Dbpedia-gsoc mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dbpedia-gsoc
