Who are your tools intended for?
Our tools are intended for experts in charge of quality, safety and more generally experience feedback.
How were your tools developed?
Our tools (PLUS, isChecker and services associated with ECCAIRS) are the result of a close collaboration with experts in the fields of safety and aeronautics. We designed these tools to respond to their needs directly related to the situations encountered in their daily activity.
Can PLUS be used to create statistics?
The databases analysed by PLUS are composed of textual data and more constrained data (dates, various type of category, numerical data, etc.). The environment proposed has the triple advantage of being able to give
- statistics on the text analysis,
- as well as statistics on the other types of data,
- and also cross-tab statistics allowing, for example, analysis of the interrelation between the results of the linguistic processing and values from the coded data.
PLUS equally offers numerous formats for data export enabling the results to be used in standard environments.
Is it possible to test PLUS?
How can I be sure that PLUS is what I need?
You need PLUS if you collect and store large quantities of natural language data and seek to explore all the valuable information it contains. When you confirm your interest to work with us, the first step is to understand your needs in order to give the best possible answer to your issues and expectations. An initial analysis of your needs is carried out, usually by visiting your company (you may prefer to meet us in our Toulouse office). Next, if it is appropriate, we will suggest a demonstration, i.e. access to PLUS with your data for a given time period.
What can PLUS do regarding well-structured databases?
Structuring a database is about storing information in an existing model, which implies that we have a precise idea of the type of information we want to keep and therefore what we are looking for. The structure models the given data and potentially puts aside relevant but unanticipated information. The structuring of information in a model, however well-defined it may be, means having to enter information into a pre-defined mould which risks making a biased analysis. If this type of practice is generally useful for gathering information together, create statistics, or exchange information with partners, it nevertheless remains that the process of categorisation is time-consuming for the experts responsible for this work and there is the constant risk that it is not done in the same way by the different analysts. PLUS represents a complimentary tool to these techniques. By using as a base the linguistic analysis of the texts contained in the experience feedback data provided by the frontline actors and having undergone no transformation, it enables to return to the source of the information : the facts described by the firstline actors. By analysing this information and by giving access in a simple way to the results obtained, it enables the experts to save time and be sure of the reliability.
Once the decision has been made to use PLUS, how long does it take to make it available to our experts?
PLUS can be set up very quickly; installation is easy and does not have an impact on the time needed for implementation. However, once the decision to use PLUS is made, it should be remembered that there is a phase of application configuration. This consists especially of defining the perimeter for action (the analysed data) and the application parameters in order to meet the expert’s needs (the functions). The length of this stage therefore largely depends on the availability of the experts in order to be able to discuss with them their problems.
Is there a minimum amount of data required for the PLUS environment?
Strictly speaking, there is no minimum amount of data required. However, it is important to note that the greater the amount of data, the more interesting will be the analysis: the correlations made will be more “solid” and the outcoming signals stronger. It is therefore preferable to have a sufficient amount of data. We are currently working on databases ranging from hundreds to several hundred thousand texts. In certain cases, only a test can give us a satisfactory answer to this question.
Does PLUS highlight the “precursors” or weak signals?
Ever since Heinrich and his famous “pyramid” of the 1930s which highlighted the relationship between the large number of incidents occurring in a company having little consequence (ex a handling error corrected by the operator himself or his colleagues, a deviation from procedure leading to only a slight reduction in quality, the loss of information resulting in a short delay, and the small number of serious or fatal accidents, the persons in charge of safety are often of the opinion that it is by dealing with the unserious incidents and reducing their occurrence that the number of serious accidents will be reduced slightly or even completely. We know however that this is not correct. The incidents are not all weak signals for more serious accidents…But the opposite is not true: amongst the minor incidents, SOME have special significance in terms of safety. The challenge is to identify them amongst the hundreds or thousands of reports. For those in charge of safety, PLUS provides the means of reaching this objective: alert in the event of an increase in the frequency of a particular type of incident described in natural language, alert when this type of incident is discovered in other areas where it was previously unknown, study of a collection of similar incidents (correlation), incidents not being linked to any other already reported and therefore considered as new, etc.
What distinguishes a tool like PLUS from the first textual static tools developed in the 1990s?
At the end of the 1980s appeared in France and throughout the world the first software programmes for processing textual data, the result of a collaboration between teams of linguists and mathematicians specialised in data analysis and initiated in France by Jean-Paul Benzécri. These programmes counted the frequency of words, calculated the number of cooccurrences and oppositions. Today, Natural Language Processing goes much further, by carrying out a complete linguistic analysis of the texts thanks to operations which enable to identify and remove any ambiguity (for example a phrase in which two words have the same spelling but completely different meanings), define the role of the words and group of words in a sentence (subject, verb…), define complex phrases , etc. By using the latest techniques in linguistic analysis of languages PLUS processes organisations/companies’ texts in and efficient and robust way in order to produce reliable experience feedback (safety, quality, trials, customer returns…)
Does the PLUS environment enable to particularize the functions of text mining of data to my domain?
Yes, the functions such as the filters for building queries for the Search are customised to your domain. More complex functions such as the Dimensions enable to model the centres of interest depending on your needs (for example identify the impact of fatigue in operator intervention reports)
In the description of the PLUS environment you talk about Dimensions. What does that refer to?
A Dimension is a model which allows to identify a particular aspect in a text. Here is an example which can explain much better than a long paragraph: in reports written by staff working with patients in an old people’s home you wish to identify everything relating to the arduous nature of their work. We set up a model representing the arduousness of their situations that we call a Dimension. This Dimension enables to organise, to classify, to sort…. the different reports according to the chosen criteria in order to analyse them.
To which area of activity does your PLUS environment apply?
The main application area currently concerns safety in the aeronautical sector due to the importance of databases and the need to exploit their contents. Other areas where the application has been used concern the medical sector (the analysis of medicotechnical risks), the oil industry (maintenance interventions in the refineries, the rail industry (operating incidents), etc.)
Generally speaking, all sectors of activity produce documents, reports, intervention reports, etc. and are all potential users of this application.
Which results does isChecker provide to the expert?
isChecker enables to make an automatic analysis of a whole procedure as well as of a large volume of documents (several thousand if necessary). At the end of the process, isChecker shows the results of the analysis in two ways: a report of each procedure one by one giving details of the variations from the reference rules (with for each variation, the rule which was not respected) and/or the comments inserted in the Word file containing the procedure at the relevant place where the variation was found. These two presentations enable the persons concerned to quickly correct the variations. At the same time isChecker also gives a global analysis of the results obtained.
What sort of information can isChecker verify?
There are several types of verification that can be done with isChecker. We can not only identify the variations of form (font, font size, format, etc.) but also structural differences (respect of the document layout, the presence of a given paragraph, etc.), differences in formulation (the use of unauthorised acronyms, ill-adapted grammatical forms, etc.) or the variations in the contents (the absence of a type of information expected for example). Do not hesitate to tell us your needs…
Which rules does isChecker use to function?
The rules that are used to carry out the verifications are customizable. Generally, we implement our client’s writing guide according to his needs. We can also complete the rules by adding standard rules of complexity analysis, vagueness markers, etc. All this is defined with you.
How much time does isChecker need to analyse large volumes of texts?
It depends on the complexity of the documents and the number of rules to be tested, but on average it takes about 1 second per page of text.
What type of document can isChecker analyse?
isChecker can be used on any type of document built from a formalised framework with writing rules. Most frequently, we work on operator procedures, operator guidelines…
Does the use of isChecker guarantee that there will be no more accidents?
The quality of procedures is essential for controlling operations and their results, both in terms of quality and safety. But even the best procedure cannot guarantee the success of an operation (from a safety or a quality point of view), if the design of the technical systems is ill-adapted to the users (who tend to lower their vigilance, make mistakes, have difficulty in following to the letter a procedure which they don’t consider to be useful…). Above all, isChecker sets out to validate the “internal” quality of the procedure, (respect of the form, absence of vagueness markers, of complexity markers, coherence between the risks identified in the procedure and the safety recommendations…) such as is recommended in the regulations in a company or a trade, or organisations (INCOSE…). But if the author of the procedure writes ‘open valve 3002 ‘ instead of valve 2003, isChecker will check this instruction from that angle only if a rule has been included stating which equipment can be used in each procedure. isChecker cannot in any way replace the competence of the operator but can help him express his competence and share it by writing better quality procedures, in other words better adapted to particular situations encountered by the operators, to the specificities of operator understanding and reasoning to the context in which the procedures are used.
Which document formats can isChecker work on?
isChecker works on electronic documents and can use the formats Word (.doc and .docx) and .html.
What is ECCAIRS?
ECCAIRS is a reporting system (European Co-ordination Centre for Accident and Incident Reporting Systems) operated and made available by the Joint Research Centre of the European Commission. Its mission is to assist national and European transport organisations with the gathering, sharing and analysis of their different information in order to improve public transport safety. It is an extremely powerful tool, requiring a certain expertise in order to exploit it correctly and effectively which can sometimes make its use complicated. We act as support to the users or future users to overcome this complexity.
At what stage do you intervene in the installation of ECCAIRS?
We can intervene at every stage of the deployment process: from the definition of the process, to the environment installation, the taxonomy and interface customisation, etc. up to the design of integrated expert functions in the ECCAIRS environment.
I am currently migrating from ECCAIRS 4 to ECCAIRS 5, when can you intervene?
We can intervene at different stages such as the migration and adaptation of data from E4 to E5, the installation, the taxonomy and/or interface customisation, the setting up of gateways to convert data from specific formats (txt, html, xls, etc.) to the E5 format (e5f).
What type of tools do you propose to complement ECCAIRS?
Our experience of ECCAIRS enables us to propose several tools to complete the functions already present in the reporting system, such as:
- Assistance for the categorisation of flight phases, occurrence category and event type.
- Assistance for the coherence analysis based on the textual analysis of occurrences which comes in addition to the Data Quality Check tool initially developed by the EASA and available as default in ECCAIRS 5.
- Assistance in the analysis of occurrences using similarity calculations
- A module for the conversion of data into the ECCAIRS 5 format (.e5f)
All the above propositions are available:
- As an add-in: the function is used in the ECCAIRS Browser to process selected occurrences
- As a standalone ECCAIRS 5 tool: the function is accessed by using an external tool linked to ECCAIRS and which therefore only requires having an instance of ECCAIRS 5 installed. It is possible to process a set of occurrences whether they are stored in an ECCAIRS 5 database or an e5f file.