An important precondition for realising the goal of a semantic web is the ability to annotate web resources with semantic information. In order to carry out this task, users need appropriate representation languages, ontologies, and support tools. In this paper we present MnM, an annotation tool which provides both automated and semi-automated support for annotating web pages with semantic contents. MnM integrates a web browser with an ontology editor and provides open APIs to link to ontology servers and for integrating information extraction tools. MnM can be seen as an early example of the next generation of ontology editors, being web-based, oriented to semantic markup and providing mechanisms for large-scale automatic markup of web pages.
An important pre-condition for realising the goal of the
semantic web is the ability to annotate web resources with
semantic information. In order to carry out this task, users
need appropriate knowledge representation languages,
ontologies, and support tools. The knowledge
representation language provides the semantic interlingua
for expressing knowledge precisely. RDF ([
The rest of the paper is organised as follows: in the next section we will show the process model underlying the design of the tool. Finally sections 3 and 4 discuss related work and re-state the main tenets and results from our research. 2
Within this work we have focused on creating a generic process model for developing semantically enriched web content. The component tools which are used in MnM are ontology servers, Information Extraction (IE) tools and augmented web browsers. During our initial work in this area we found that either the existing tools did not directly support the creation of semantic web content or the mapping between the tasks to be carried out and the toolset was non-trivial. Hence, within MnM, we adopted a generic process model, which can be easily understood by web developers who are not necessarily expert ontology engineers or human language technology experts.
Another key feature of our process model is that it is generic with respect to the specific ontology server and IE technologies used.
There are five main activities supported by MnM: • • • •
marked up corpus to learn the extraction rules.
•
In this activity the user browses a library of knowledge models which sit on a web based ontology server. The user can see an overview of the existing models and can select which one to focus on (i.e., which ontology to use to initiate the markup process). Within a selected ontology the user can browse the existing items - for example the classes and instances. Items within an ontology can be selected as the starting point for selecting an IE mechanism. More specifically, the selected class forms the basis for a template which will eventually be matched against a corpus of documents and instantiated in the extraction activity.
Once a class has been selected a training corpus of manually marked up pages needs to be created. Here the user views appropriate documents within MnM’s built-in web browser and annotates segments of text using the tags based on the class’s slot as given in the ontology (i.e., ontology driven mark-up). As the text is selected MnM inserts the relevant XML tags into the document.
MnM integrates web browsing, ontology browsing and IE development. It does not have a built-in IE tool but provides a plug-in interface which allows the integration of IE tools easily.
In a previous version of our MnM we integrated Marmot,
Badger and Crystal from the University of Massachusetts
([
Amilcare is designed to support active annotation of documents. It performs IE by enriching texts with XML annotations. To use Amilcare in a new domain the user simply has to manually annotate a training set of documents. No knowledge of Natural Language Technologies is necessary.
Amilcare is designed to accommodate the needs of
different user types. While naïve users can build new
applications without delving into the complexity of Human
Language Technology, IE experts are provided with a
number of facilities for tuning the final application.
Induced rules can be inspected, monitored and edited to
obtain some additional accuracy, if required. The interface
also allows precision (P) and recall (R) to be balanced.
The system can be run on an annotated unseen corpus and
users are presented with statistics on accuracy, together
with details on correct matches and mistakes. Retuning the
P&R balance does not generally require major retraining,
facilities for inspecting the effect of different P&R balances
are provided. Although the current interface for balancing
P&R is designed for IE experts, a future version will
provide support for naive users ([
At the start of the learning phase Amilcare preprocesses
texts using Annie, the shallow IE system included in the
Gate package ([
Amilcare then induces rules for information extraction. The
learning system is based on LP2, a covering algorithm for
supervised learning of IE rules based on Lazy-NLP ([
All the generalizations are tested in parallel by using a
variant of the AQ algorithm ([
Correction rules are identical to tagging rules, but (1) their patterns also match the tags inserted by the tagging rules and (2) their actions shift misplaced tags rather than adding new ones. The output of the training phase is a collection of rules for IE that are associated with the specific scenario.
MnM provides various mechanisms for selecting a test corpus and distinguish this from a training corpora. The user can manually select training and test corpora and these can be in the form of local files or on the web. In addition, it is also possible to simply select a corpus (either locally or on the web) and let the system to create, to test and training corpora randomly.
After the training phase Amilcare has a library of induced
rules which can be used to extract information from texts.
When working in extraction mode, Amilcare receives as
input a (collection of) text(s) with the associated scenario –
scenario is the set of tags that the user will insert in the
training corpora- (including the rules induced during the
training phase). It preprocesses the text(s) by using Annie
and then it applies its rules and returns the original text
with the added annotations. The Gate annotation schema is
used for annotation ([
Once that is done the information extracted is presented to the user for approval. Then the extracted information is sent to the ontology server which will populate the selected ontology.
During the population step the IE mechanism fills predefined slots associated with an extraction template. Each template consists of slots of a particular class as defined in the selected ontology, for instance, the class visiting-a-place-or-people has the slots: visitor, place, etc.
Our goal is to automatically fill as many slots as possible. However, some of the slots may still require manual intervention. There are several reasons for this problem: • there is information that is not contained in the text, • none of the rules from our IE libraries match with the sentence that might provide the information (incomplete set of rules). This means that the learning phase needs to be tuned.
The extracted information is also validated using the ontology. This is possible because each slot in each class of the ontology has a type associated with it. Therefore, extracted information which does not match the type definition of the slot in the ontology can be highlighted as incorrect.
A number of annotation tools for producing semantic
markup exist. The most interesting of these are Annotea
([
While both MnM and OntoMat are very similar they illustrate a slight difference of emphasis in providing tools for the Semantic Web. While OntoMat adopts the philosophy that the markup which indicates the knowledge content of a web resources should be included as part of that resource, MnM’s annotations are stored both as markup on a page and as items in a knowledge base held on the WebOnto combined ontology and knowledge base server. 4
In this paper we have described MnM, an ontology-based annotation tool which provides both automated and semiautomated support for annotating web pages with semantic contents. The first prototype of the system has now been completed and tested with both Amilcare and the UMass set of tools. The early results are encouraging in terms of the quality and robustness of our current implementation, however, there is clearly a lot more work needed to make this technology easy to use for our target user base (people who are neither experts in language technologies nor 'power knowledge engineers'). In particular, all the activities associated with automated markup tend to be very sensitive to the quality of markup and to the appropriateness of the chosen corpora. Amilcare already attempts to address some of these issues through its adaptive mechanisms, however, more work is needed in this area. In addition, we also plan to do more work on the user interface, in particular with respect to the integration of markup, ontology browsing and the 'semantic navigation' of web pages. Currently, ontology and web browsing are integrated with respect to contents annotation, but ontologies do not inform the web browsing component of MnM directly. Our vision for the semantic web is one in which new forms of 'conceptual navigation' will emerge, where association between resources will be semantic as well as hypertextual. We plan to experiment with these ideas and extend the interface of MnM to support novel, markup-driven forms of web browsing, as well as the standard HTML based ones.
This work was funded by the Advanced Knowledge Technologies (AKT) Interdisciplinary Research Collaboration (IRC), which is sponsored by the UK Engineering and Physical Sciences Research Council under grant number GR/N15764/01. The AKT IRC comprises the Universities of Aberdeen, Edinburgh, Sheffield, Southampton and the Open University. The authors would like to thank Maruf Hassan and Simon Buckingham Shum for their invaluable help in reviewing the first draft of this paper.