MediaIBIS bug reporting and suggestions

From KeyToNature
Jump to: navigation, search

This page is intended for project partners:

  • to report their results in using the mediaIBIS metadata search engine (see Test reporting section bellow) and
  • to bring suggestions on how to improve the application (see Suggestion section bellow).

Test reporting

When reporting possible bugs in your test, please let us know:


Your Name

  • mediaIBIS version: Eg. ID-Tools Search or General Media Search and interface language: Eg. EN, RO, etc
  • description of the issue and, if the case, a suggestion how to fix it: ...
  • how often you encounter the issue (by following the same steps when using the application the bug could appear always, randomly or only once): ...
  • the steps to reproduce the bug: ...
  • if the application fails altogether: your operating system, browser name and version and flash plugin version.


Pier Luigi Nimis

General media search:

  • I tried with ‘Fagus’, obtaining 60 results. Then I tried with ‘Fagus sylvatica’ and I was surprised to find many more (241) results. The reason is that by default the machine looks for ‘fagus’ OR ‘sylvatica’. I would get more or less the same results by searching ‘sylvatica’ alone. This is rather annoying. Then I have found an instruction (in the manual) telling me that I should have put my text between double quotation marks. I think that many users will never read the manual, and that it would be better to put some very simple instructions in the query interface itself. In any case, I think it would be best to put as default the AND operator.
Gregor: One caution about always searching for quoted "Fagus sylvatica": this may also lead to cases, where relevant material is not found. The logic of the full text search is too give too many results, but to give the more relevant first. We might have to define whether some use cases intend exhaustive searches (paging all 13 pages to 241 to be sure all has been seen). This is currently not anticipated. User interface recommendations are highly welcome though: space in the search interface is very limited (1024x786 screens!) - how shall we explain the expert search options?
  • Sometimes it would be very useful to look not for words but for strings of text (e. g. “gus sylva”). This feature is not implemented (in google it is).
Gregor: I believe we cannot support this, we are bound to the abilities of the full text index (Lucene). Also, how do you get Google to search for “gus sylva” and find Fagus sylvatica? Google will find some hits, because the documents are broken (a hyphenated line start as "gus") but I could not get google to search instring.
  • Some pictures (from ETI) of Fagus appear as offline. However, as soon as I go their metadata, I find a link which brings me to a picture.
Gregor: We must check this. Andrei: Is the availability correctly specified by ETI?
Mircea: It is true. It needs to be corrected by looking not only to "Availability", but also in "Best Quality".
Lia: Some pictures from ETI still remains offline just because they are not in an appropriate format, just an background image of a html page on ETI site. Some of those collection are already re submitted by ETI, so the situation are getting better. It's not the mediaIBIS problem. It's a metadata quality problem. This is visible only through mediaIBIS and so we have a very good feedback.

About UNITS metadata - see Algae_(UNITS), Fungi_(UNITS): many Best_Quality_URI are wrong, being of the form: "picture.jpg.jpg". It's seems to be an export error, many times reported by our ingest tool. I should remove this bug by programming a special sequence.

  • On the whole, I like very much the outline. In 2 days I’ll spend a whole afternoon for going on with the test.
Many thanks for the testing!

Lia Veja

  • mediaIBIS version: General Media Search, Language: EN
  • description of the issue: Why Best Quality are displayed twice?

Best Quality: http://212.201.100.117/storage/Fungi/JKI-IZ-DIA/web/5478.jpg; http://212.201.100.117/storage/Fungi/JKI-IZ-DIA/web/5478.jpg

Alessandro Mazzetti

  • Find_media generally works, but it is very slow...
  • Find_identification_tools is difficult to use because you are required to know the name of the tool.
Mircea: Not necessarily. You have bellow of the text box to search 4 buttons from which you may select: the availability, the language, the organism grup, the platform - all in a combined form. As an example you can search: ALL Id tools (no text in the box), where Availability = Online AND Language = Italian.
  • Searching for "Salvia" nothing is found, but searching for "scuola" you get a tool which identifies "Salvia".
Mircea: It appears like this because "Salvia" is not a keyword in the *** metadata *** description of the tool, but it is something that it is inside the tool. The search engine doesn't analyse the content of the Id tool, but its metadata.
  • Searching for "cupressus" you get a pdf document but not the tool that identifies the "cupressus".
Gregor: To separate problems: This is the responsibility of the data providers. Or have you checked that the tool that you know is identifying Salvia informs about this (e.g. Salvia is in taxon list)? Many providers to not provide a taxon list, which means their tools cannot be found for this type of query. Providers can update their metadata any time by reuploading modified metadata to keytonature.
Mircea: See my comments above. "Cupressus" is the ** title ** of the tool.
  • Sometimes a Database error occurs, see attached screenshot.
Gregor: Our problem for sure!
Mircea: How many times in a working session do you encounter this? Is it somwthing regular? (at each working session?).
Alessandro: No, it is occasional, it seems having to do with rapid sequence of searches

Marina Ferrer

  • On the first place, I would like to congratulate the creators for such am important step forward. Below you can find some remarks:

Media IBIS general media search:

  • I have changed the metadata for images (specifically the license statement) and I can not see now the change at the engine. Does it take a while? Are there automatic revisions periodically?
  • Similar to what Pier Luigi wrote: I type "Pinus nigra" for images and I obtain about 300 items. By typing "nigra" alone I obtain only 198, and if I type "pinus" alone I obtain only 153, which all together doesn't seem to make much sense…Besides, by typing "nigra" alone I don't obtain any image of "Pinus nigra"…
Gregor: How can we solve this? The fundamental problem is that the only hit in a full text search may be LESS than the combined search terms. For example the image may be labeled "P. nigra". For this reason, full text search usually uses a strategy to find the OR-combination, but always put the AND-combination first in results (it is even slightly better than that, having addition relevance and adjacency criteria). If the Pinus AND nigra do not come first it is an error. Interestingly, although Google uses the same principles and similar technology, they "create" reasonable numbers: Pinus 1,240,000, nigra 1,700,000, Pinus nigra: 1,070,000. It then only allows access to the first 750 hits, which brings it generally into safe water. Interesting social strategy to avoid the confusion both Marina and PL are describing. Still do we want to rather have users not find anything?
Mircea: I have noticed that for "Pinus nigra" search, the first 30 images have all in the name "pinus nigra". Starting with the 31st image ("sambucus nigra"), the names are "XXXX nigra" up to image 306, where start the name "Pinus YYY" and ending with "Pinus" Id tool (RJB).
Mircea: Marina, about the "Besides, by typing "nigra" alone I don't obtain any image of "Pinus nigra"…" --> the images are there, but not starting on the first display page. They start at the image 46.
  • I have had difficulties for viewing the videos, to some of them I only obtain a question mark symbol. I wonder if it has anything to do with the engine structure or not…
Mircea: As I see most of the videos are encoded and you would need a Video Player for this (in the cases of ETI's videos the system requires QuickTime Video Player). If some of you encounter problesm also with this player, please report.

Media IBIS identification tools search:

  • I have looked for "identification tools" of "vascular plants" in "Spanish" (not typing any string) "all available resources", and I obtain no results (and I know there are). Typing a concrete genus (Quercus) shows no results either…
(Gregor:) can you check this, Andrei? Marina: can you show the metadata page on the wiki for your expected results?
Mircea: For "Quercus" I get one ID key at the address: http://www.floraiberica.es/floraiberica/texto/pdfs/02_041_03_Quercus.pdf.
Andrei: I tried the combination above and it returned 493 results... Maybe there were some issues with the Gsearch index again?
Lia: Identification tools of "vascular plants" in "Spanish" returned 7 results. Is that correct?
  • About the query interface, how do we get it in our language? Is it not possible yet? It is translated, right?
Mircea: At the link http://www.keytonature.eu/wiki/MediaIBIS_first_general_testing, see the table and go to the raw "Spanish".
Marina: I choose "Spanish" but I only get the headline in Spanish, the rest keeps in English...

Thomas Putz

First of all thank you for the nice work you did with this search tool

General impressions

  • very nice interface
  • quite easy to understand and choose the options
  • One hint: Austrians and people from Switzerland react a bit "strange", if they see only the German flag to choose the German language option. Better to use a combination of 3 flags to indicate the German language version.


I was logged in during the test.


Media IBIS identification tools search: Language German (all other Settings ""beliebig")

  • first I tried to enter "Strauch" as shown as hint for the search - no results
  • then I tried other terms in German like "Baum", "Tanne" - no results
  • searching for a special genus like "Qercus" does not show any results neither
  • sometimes I got an error #2032 instead of "Es wurden keine passenden Ergebnisse gefunden."
  • a search for "Austria", "Österreich" or "Graz" did not show any results
  • at last I tried search terms in English (language setting still German) and got results
    • "vascular plants" => 2 results
    • "tree" => 0 results
    • "animals" => 0 results
    • "Lichenes" => 0 results

(Andrei) At the time I tested the application there were only 3 identification tools in German language.

Media IBIS general media search: Language German (all other settings ""beliebig")

  • first I tried to enter "Strauch" as shown as hint for the search - no results and in the message was stated that I should change the settings in the "Verfügbarkeit-Menü" - setting is "beliebig"?
  • then I tried other terms in German like "Baum", "Tanne" - no results
  • searching for a special genus like "Qercus" does not show any results neither
  • a search for "Austria", "Österreich" or "Graz" did not show any results
  • at last I tried search terms in English (language setting still German) and got results
    • "vascular plants" => 26 results
    • "tree" => 106 results
    • "animals" => 7 results
    • "Lichenes" => 196 results
  • Very nice representation of the results - both options work very well
  • only one remark: the mouse other zoom works with some pictures not as an enlargement - it zooms in as well (i.g. armelia / Evernia – Evernia)
  • A search for "armelia" shows no results, a search for "Evernia" shows 11

Verónica Donoso

Nice work so far! Below you'll find some of my comments. i hope they can be useful for some improvements.

  • When you click on the “visit resource” button you would expect to find more information than simply getting the current picture enlarged as it currently seems to be the case (at least in all the examples I tried). Thus, it would be advisable to either change the name to “enlarge picture” or add more resources.
    • (Andrei): We can't change the name to "enlarge picture" because we use the same icon of "visit resource" for all online data. Enlarging would also be a improper term, because we just point the user to the original data, and in the case of pictures, they will automatically be open by the browser.
  • When you click on “resources links” (not in the icon visit resources) the previously mentioned problem occurs once again, i.e. users expect to find other resources and they only get an enlarged version of the picture they are currently looking at. Besides, having the same information presented via two different labels is redundant, inconsistent and confusing for users. On the one hand, they get the same information twice (via an icon and via the link in “resource link” section). On the other hand, they get access to the same information via two different names.
    • (Andrei): The section "resource links" contains all links related to the resource. This means that we can have multiple links to multiple qualities of an image. The "visit resource" link will always point to the highest quality available, but the user still has the option to view lower qualities.
  • Why do you have a “technical section” and an “additional metadata” section? Couldn’t all that information be placed under the same heading? Or is the technical information different than the metadata? If it is so, then 2 sections would be justifiable. However, currently both sections seem to overlap.
    • (Andrei): The "additional metadata" section has been created for metadata fields that have not been categorized or translated. There is not connection between "technical section" and "additional section"
  • When you are looking at specific information from “Redirected from Find media (mediaIBIS) you get to an intermediate page with no content but a link “Find images and other media (mediaIBIS)”. If you click on the link, just then you’re redirected to the actual mediaIBIS search engine. This last step is unnecessary. Users should immediately be redirected to the mediaIBIs without having to click on unnecessary screens.
    • (Andrei) this issue will be solved when mediaIBIS will be directly accessible from the homepage.
  • A “back” button is missing from the results page.
    • (Andrei): where should the "back" button point? The users have the option to return to the main search screen via the New/Revise Query Links.
  • MediaIBIS finds resources for terms such as “rosa”, but not for” rosas” (plural). If a user types the plural of a plant the results of the singular should also be displayed.
    • (Andrei): Users have the ability to use the tilde ~ to perform lazy searches. E.g. rose~ will search for rose, roze, rosea, etc.
  • “Search string” vs. “new query”. These two different terms are used to refer to the same act of looking for information. The terms used should be consistent across the interface so that users can easily recognize and recall them.
    • (Andrei): The terms are different, but they address different things. "Search string" refers only to the keywords the user enters in the text field and "query" is represents all user entered and/selected options (keywords +/- search buttons).
  • The sound icon does not work. It’s necessary to click on the link, for example http://www2.pms-lj.si/cdbarje/Pteronemobius-heydenii.mp3.
    • (Andrei): The sound icon is there only to let the user know visually what type of data he is dealing with. In case the resource is online, to access it he would have to use the "Visit resource" icon. Similar icons are available for other datatypes too.
  • What’s the difference between “digital” and online resources? I don’t think both terms are really mutually exclusive, but they should be.
  • Using mediaIBIS in Mozilla creates some problems. For instance, when no pictures are available online (e.g. when you type in a term such as “rosa”) the results of the species without a picture (in Internet Explorer you get a box with the message “image is offline” that can be clicked in order to get the metadata) are not shown and it is not possible to get any general or technical information nor metadata.
    • (Andrei) What version of Mozilla were you using? I've tested with Mozilla 3.5.7 and had no such issues.
  • In internet explorer, when no pictures are available it is still possible to get the general/technical information plus the metadata. However, what is not very logical is the fact that even when users get the message “the image is offline” it is possible, by clicking on the “visit resource” button, to get online access to the picture.
    • (Andrei) This is due to an issue with the metadata.
  • The options “New Query”, “Revise Query”, “Switch to: Table View” are not easily seen, probably because the page is already crowded with lots of information. Make these options more visible and/or place them in a more visible place.
    • (Andrei) Due to limited space and other complications the options above will remain how they are.
  • When looking for available resources, are results presented in any logical way in the default option (i.e. when not displayed on the table)? As far as I can see results are not displayed alphabetically and in some cases the first resources displayed are those where no pictures are available. This may seem confusing for users and may give them the wrong impression that no further information is available, especially if they are using Mozilla where no further information is available. Thus, I would suggest displaying the results with pictures first.
    • (Andrei) The results come back from Gsearch ordered by relevance, high to low. The solution adopted to display online images first was launching mediaIBIS with "Online (free)" being preselected from the "Availability" menu. When it still goes wrong, it means that there are issues with some of the metadata.
  • There should be a more user-friendly way to access the media engine from the results page. Once you are on the results page, there’s a “previous” and a “next” button, but not a “back” button to easily return to the “(new) query page”. A Back button that takes users to the previous query page should be added.
    • (Andrei) a "Back" button would be redundant as the user has the "New/Revise" Query Options.
  • There are some more serious problems with the search itself. For instance if you type in the term “willow” the first result you get is “Rhytisma salicinum”, i.e. a type of fungi that grows in a willow. The term willow only appears in the general information (within description) “Rhytisma salicinum: Black apothecium formed in leaf tissue of a willow (Salix cinerea)”. However the search engine still shows the “Rhytisma salicinum” as a relevant result. The same happens if you type in “dog” or other vernacular terms. This problem requires a more complex solution as it has to do with the way of searching for information within the database. Gregor, Mircea, do you have any suggestions to improve this?
    • (Andrei) Again this is due to the Gsearch scoring system. There is no current fix for this issue.
  • http://www.keytonature.eu/wiki/Find_media_(mediaIBIS) –It’s not a good idea to add () to the name of the link

Suggestions

For any kind of suggestions related to mediaIBIS search engine, please use the table below.

Name Topic Suggestion
Andrei Homodi User interface request Please redo the navigation for the results. I would like to see the results from the last page.
Mircea Giurgiu Page display I would like to navigate directly to the "Last page", "First page", or on a specific page number, when I browse through resources.
Mircea Giurgiu Display of images in grid mode Because you display 3 images / row, it would be good that a page contains a multiple number of 3 (e.g. 21 images / page, instead of 20 - as it is now: 24.11.09)
Thomas Putz flags to indicate languages use a combination of the German, Swiss and Austrian flag to indicate the German language version
Thomas Putz Search hints IMHO it's very important that the terms shown as hints for the search lead to results
v
x