Searcher
Vol. 9 No. 5 May 2001
THE BETTER MOUSETRAP
Derwent on Delphion: An Experiment
by Nancy Lambert Senior Information Analyst
Chevron Business and Real Estate Services Company (CBRES)
Table of Contents Previous Issues Subscribe Now! ITI Home
The Derwent World Patents Index is unquestionably the most important resource available to patent information professionals for subject-searching international patents. In chemical and polymer areas, its generic search capabilities complement Chemical Abstracts' precise Registry Number searching; in non-chemical areas, it provides unique subject access.

However, Derwent's indexing comes at a high price. To obtain access to "deep indexing" (the in-depth chemical and polymer indexing, the broader chemical indexing system called manual codes, and some related chemical subject access), companies must pay a substantial up-front subscription. A subscription to access all deep indexing in all subject areas currently runs over $100,000 per year.

This has suddenly changed. Delphion Inc. [http://www.delphion.com; formerly the IBM Intellectual Property Network] announced on February 26th that it has loaded the Derwent World Patents Index on its Internet site and that anyone, subscriber or non-subscriber, may now search all of Derwent's indexing. Derwent sees this as an experiment in extending Derwent access outside its subscriber base. But what it has loaded on Delphion raises many questions.
 

First Look at Delphion's Derwent
Let's take a look at Derwent on Delphion. First you go to the site and log in (see Figure 1). If you haven't registered yet, you must do that first, but it's free and fast. When you have logged in, you go to a page that gives you a number of choices, including the other Delphion search pages: U.S., European, Japanese, and PCT (Patent Cooperation Treaty, or World) patents, plus Inpadoc, the International Patent Documentation Service (see Figure 2). From here you can go to Derwent via at least three routes: Click on "Derwent" in the menu on the left side of the page; click on "Derwent World Patents Index" in the menu on the bottom of the page; or click on "Derwent" in the submenu under "IP Search" on the top of the page. (You can access Derwent this third way from any Delphion page.)

On the opening Derwent page (see Figure 3), you notice immediately that you must submit a billing activation form in order to access the database. If you do not have a Delphion corporate account, you must either set one up or provide a credit-card number to access the paid areas of the site, including Derwent. If you order the billing activation as a non-Derwent subscriber, you can access Derwent data immediately. If you order billing activation as a Derwent subscriber, Derwent's verification will take several days; but you can still access Derwent data immediately at non-subscriber rates. Anyone may access the free demo searches, which operate on 1 week of Derwent data.

Both demo and full searches (see Figure 4) include a choice of Boolean and Advanced search pages, plus a patent number/accession number lookup page. The Boolean search page (see Figure 5) uses a typical Internet search form: It includes several boxes into which you key search terms. Each one lets you search "all fields" or choose a specific field. The boxes are default-connected with Boolean AND logic, but you can change this to Boolean OR or NOT.

Under the search boxes, you will find a line that reads "Chemical and manual code fields (Deep Indexing)," followed by a link named "Expand++." When you click on this, you go to another page (see Figure 6) that has the initial search term boxes followed by a series of boxes that let you choose from manual codes (the default), a whole variety of chemical codes, polymer indexing, and polymer multipunch codes. This page does not give you time ranges for any of the codes.

The Advanced search page (see Figure 7) starts with an "Any field" box that lets you put in a multi-field, multi-level Boolean search of your own choosing. (You click on "query language," in the menu down the left side of the page for information on the Verity search language syntax.) Or, you may choose to put a free-text subject search into this box and then further refine your search by limiting to inventors, patent assignees, any of a number of date ranges (not available on the demo search), International Patent Classes (IPCs), Derwent classes, and various text fields, including Derwent title and abstract. (Remember your British spellings!) (See Figure 8.) The system does not give you a choice of Boolean operators between these fields or tell you what it is, but a bit of experimentation determined that it's AND logic. Again, the Advanced search page lets you click on "Expand++" to get to a search form for the deep indexing; here you see the date ranges available for the different kinds of codes. (More on this later.) This page has at least two typos: It lists polymer multipunch codes (which started in 1966) as 1986 to 1994; and it states that the chemical codes (which started in 1963) are 1981 to date only. In both cases, sample searches belied this, producing records from the start of the indexing systems.

Note that polymer indexing applies only to Derwent section A, polymer chemistry; however, the chemical codes apply to sections B (pharmaceuticals), C (agrochemicals), and E (general chemistry)); additional sets of chemical codes apply to dyes, steroids, and drug delivery formulations and processes. On the traditional online hosts, searchers can choose to any combination of these sections with the chem codes. Delphion defaults to searching all sections. You can limit to specific sections, but documentation is not yet available on how to do this.
 

Pricing
From the Derwent opening page you can click on a link to pricing (see Figure 9). But be prepared for sticker shock. Pricing has no connect-time element; instead, it is based on the number of search queries submitted and items displayed. Derwent differentiates between "basic search" (no deep indexing used) and "extended search" (deep indexing used). Note that in Derwent's pricing definition, deep indexing includes all the manual codes, both the chemical, which require subscriber access online, and the electrical, which have been available to non-subscribers for some years now. As of March 15th, prices are as follows:

  • Basic search: $5 for subscribers, $8 for non-subscribers.
  • Extended search: $59 for subscribers and non-subscribers alike. (This has some implications that I discuss later.)
  • Short-format display: Between 20 and 50 cents per record, depending on your company's subscription status. The system defaults to 10 records per page, but you can change this.
  • Long-format display: Between $1.70 and $4.50 per record depending on subscription status.
The long-format display prices seem at first just slightly more than traditional online full-display prices. However, Delphion displays do not include some of the segments in the new abstract formats that Derwent introduced in early 1999, most significantly the technology focus (details of interest to scientists and engineers). Nor does Delphion give Derwent subscribers the option to include the extension abstracts (supplementary chemical and biological detail) available for an extra charge on the traditional online hosts.

Keep in mind that the technology focus and extension abstracts are not there for searching, either. A search of patents assigned to Exxon mentioning metallocenes, with accession years 1999-date, produced 34 records on DWPX (Derwent with extension abstracts) on Questel-Orbit, only 26 records on Derwent on Delphion. (This test was complicated by the fact that Delphion should in fact have caught several [but not all] of the records it missed. Turns out that several weeks' patent assignee data was missing from Delphion. This too should be corrected quickly.)
 

Linking to Delphion Displays
On both the Boolean and Advanced search pages, you can tailor your short-format display to include the Derwent title plus any or all of the Derwent update code, Derwent title, patent number, issue date, inventor, assignee, and IPC. Then, when you do a search, your short-format search results will show all the fields you chose (see Figure 10). Each reference starts with the Derwent accession number highlighted. When you click on an accession number, you go to the full Derwent record for that reference (see Figure 11).

At this point you can start to take full advantage of Delphion's strong points: its linking and display capabilities. For instance, suppose a Derwent record that you retrieved in a search shows a patent number from one or more of the patenting authorities that Delphion covers: the United States Patent Office (US), European Patent Office (EP), Patent Cooperation Treaty (WO), and Japanese Patent Office (JP). If you click on the patent number, the system takes you to the Delphion record for that patent (see Figure 12 and Figure 13). From here you can order a full copy of the patent. From here, too, you have further links to more information at no additional cost:

  • Patent images (full patents for US, EP, and WO; Patent Abstracts of Japan for JP).
  • Further information about the patenting company via "News, Profiles, Stocks, and More"
  • Older patents cited in this patent and more recent patents citing it.
  • Legal status information about the patent (for all but JP).
  • Related patents (continuations, continuations-in-part, divisionals).
  • Full patent family information, from Inpadoc.
The patenting company link for now takes you, not directly to the information, but to another site outside Delphion (Hoover's) where you must search further. It also gives you a link that lets you search the company's name in a standardized USPTO format to see a list of all that company's U.S. patents.

The family listing can be rather overwhelming at first glance (see Figure 14). Closer inspection shows that for multicountry patents (EP, WO), Delphion does not simply display the designated states, but repeats the EP or WO patent publication numbers for each country designated, with the country code tacked onto the front of the number. Delphion says this was unintentional, and it should be corrected by the time you read this.

Patent classifications (IPCs and, for U.S. patents, U.S. classifications) are also highlighted on the Delphion records. If you click on a U.S. patent class/subclass, the system searches that subclass and shows you a list of U.S. patents classified there. However, if you click on an IPC (e.g., E21B 43/263, stimulating oil production in a well by fracturing with explosives), the system searches the entire four-character IPC (E21B, everything on earth or rock drilling) rather less useful.

If you click on "Show legal status actions," you see a table of actions for US, EP, and WO patents. Positive actions are coded in green; negative actions, in red; and neutral actions, in black. At the top of the legal status table, a link appears reading, "Show all possible codes for US" (or EP, or WO). If you click this link, you get a list of legal status codes for that patenting authority, grouped by negative/positive/neutral and alphanumeric within these groups, with the brief definitions that appear in Inpadoc legal status actions.

All of this linking involves a great deal of running back and forth, starting with the short-format list of search results. Delphion could cut one level of required links by letting searchers scan the Derwent patent titles, which are fairly descriptive, and check off a subset of references to display full Derwent records. Searchers could then continue using this selected list.

One caveat (which may be obvious): In order to get from Derwent to the Delphion records and the further links, you have to display the full (high-price) Derwent records. You can't link out from the short-format display. Of course, if you include the patent number in your short-format display, you can search the Derwent basic patent directly on the Delphion Patent Number search page and display the Delphion record for free.
 

The Problems
Some years ago, Derwent and IBM tried for an alliance. This effort failed; but before it did, Derwent showed the proposed product to searchers attending that year's Derwent Subscriber Meeting. This product had limited subject access and did not include deep indexing. Some of us including me were guilty of telling Derwent that if it were to load the Derwent World Patents Index on an Internet host, it should load the whole thing.

The problem is, we hadn't thought it through. Derwent's deep indexing (except for the manual codes) is not suitable for end-user searching. And the Delphion platform is not suitable for professional intermediary searching. Although Delphion has put excellent technical work into this mounting, its Derwent product doesn't work well for either end users or professional searchers. It's doubtful at this point who will search Derwent in-depth on Delphion.

Professional searchers almost certainly won't. Even if this Derwent mounting had no other problems, we seldom use just a single patent database for any given patent search. No one database, no matter how well-indexed and expertly searched, will produce complete results; we must use all appropriate databases and take advantage of their different indexing systems. We need to do this on the traditional online hosts with their multifile/cross-file search capabilities. (Have I and others said this before? Derwent, are you listening?)

But any Internet mounting has other problems beyond the single-database issue. The Delphion platform has serious search limitations. If you want to do something approaching an online Boolean search combining multiple fields, you have to learn and use the Verity command language, which makes the online hosts' command languages pure simplicity by comparison. (Take a look at "query language" and you'll see what I mean.)

More important, you cannot create and combine Boolean sets the mainstay of the professional online searcher. The best you can do is to enter your search in one statement and then go back and refine your search if the first pass didn't get what you want. This approach is difficult for complex searches, especially when you want to combine different search parameters (text words, IPCs, Derwent codes) as synonyms or when you need to take several approaches to the same question. Furthermore, the pricing makes this impractical. As I mentioned above, a single search using any of Derwent's deep indexing currently costs $59 and refining a search query counts as another search. (Searchers may of course practice and refine their searches in the "demo" search forms before committing to the paid searches, assuming that they can extrapolate accurately from 1 week's worth of patents.)

Another issue is linking Boolean, not hypertext. Derwent's chemical codes must be linked to each other if the searcher is to stand a chance of finding patents in which the various substructures actually appear in the same compound. Derwent's current polymer indexing (applied since 1993) requires three levels of linking for maximum-precision searches. At first glance Delphion appears not to have the linking capabilities needed for proper searching of the codes. In fact, it does, within the Verity command language; but the linking is very laborious, and again Delphion has not yet produced the documentation to teach searchers how to do it.
 

Training
Another question is how on earth end users could expect to learn Derwent deep indexing, especially the chemical and polymer codes. The codes are incredibly user-hostile. Even if the end users are chemists and can make some sense of the coding sheets and manuals, they would need extensive training to learn Derwent's complex usage conventions and to understand when certain codes are and are not used. And if you make an error of commission that is, search a code that Derwent does not index for the context you want you get zero correct retrieval. Searchers need both in-depth training and frequent practice to use the codes correctly.

Derwent does not plan to support training; that is Delphion's responsibility. Delphion is working on a training module for the manual codes, but has not looked at the chemical and polymer codes yet. I suspect Delphion will be in for a shock when it does. My first training on the deep indexing took 2 full weeks. To say the least, that would be difficult to duplicate in online or video training modules.
 

Time Ranging
Both sets of codes (chemical and polymer) are further complicated by the fact that they have changed considerably and frequently over time. End users cannot possibly know all iterations of the codes and how to search them correctly for the different time ranges. In 1981, when the chemical codes underwent the last major revision, old codes were translated into new-code format and old records reloaded with the new form of indexing. So one set of chemical codes covers 1963-date, if you know how to search the different time ranges correctly.

Polymer codes, however, have three different formats for three different and somewhat overlapping time ranges: multipunch codes (1966-1994), key serials (1978-1994), and the new polymer indexing (1993-date). On the Advanced search page, Delphion has a separate box for each of these. (The Boolean search page lets you search only multipunch codes and new polymer indexing.) Given Delphion's current defaults on the Advanced search page, this is a problem. If you want to search the entire time range and therefore enter the synonymous codes into the appropriate boxes for instance, search for polyethylene by putting "P1161" into the polymer index coding box, "047 and 688" into the polymer multipunch code box, and "0232" into the polymer keyterm serial box the system combines them with Boolean AND rather than Boolean OR, and you end up only with the few hits between 1993 and 1994 indexed with all three codes. The only way to OR these codes together in one search on the Advanced search page is to search them in the "Any field" box using the Verity query language. Or, you can do the search with the implied AND operators, copy the search logic that appears on top of the first page of search results and paste it into the "Any field" search box, change the AND operators to OR, remember to remove the codes from the search boxes below, and re-run the search. (And, presumably, pay twice for the search.) Whew! You are better off searching polymer indexing on the Boolean search page, where you can change the Boolean logic between search boxes. This is something Delphion should consider for the Advanced search page also.

Incidentally, if you follow Delphion's example of entering multiple codes (which should be linked together) separated by commas, e.g., 047,688 for ethylene homopolymer or G100,K431 for benzenesulfonic acid, the system ORs them together and gives you everything indexed with either one of the codes. If you enter them separated by a space, the system appears to look for them as an exact phrase and cuts your retrieval down to near-zero. You must enter them connected with a Boolean AND. For now, the search form explains none of this.

One more small caveat: Date ranging on Derwent on Delphion can be deceptive. Keep in mind that a granted patent may issue some years after its first equivalent family member has published. Suppose, for instance, a U.S. patent issuing in 1998 was first filed in 1995 and first published as a PCT application in 1996. Online on Questel-Orbit, you can choose several ways of time ranging. If you search patents 1998-date ranging by accession year, you get only Derwent records whose earliest patent family members appeared on Derwent since the start of 1998 and you don't get your 1998 U.S. patent whose PCT counterpart appeared in 1996. If, on the other hand, you search patents 1998-date ranging by patent date, you search the publication dates of all patent family members and retrieve all records in which any patent family member appeared 1998 to date. This will include a number of Derwent records with much earlier accession dates and it will include your 1998 U.S. patent.

This is not the case on Delphion. Delphion date-ranges on the publication date of the Derwent basic patent (the first one Derwent picks up in a patent record). Issue dates of later family members can be searched directly but not date-ranged. So the caveat is, if you want to see all patents (or a specific patent) issued/published in a given time range, you need to look back at least 3-4 years earlier to be sure of including them.
 

Bad Scenarios
So Derwent's polymer and chemical coding are unsuitable for end users and difficult even for professional intermediaries to search on Delphion. This situation is further complicated because Delphion does not provide security access privileges. For now, companies who subscribe to Delphion in order to give their end users access to patent copies and free searching cannot block end-user access to the definitely-not-free Derwent searching. Nor can Delphion now separate patent copy charges from Derwent search charges. So imagine a multitude of end users stumbling onto Derwent on Delphion, using it ineffectively, spending a fortune (and busting their library's budget in the process), and deciding never to return. This wouldn't be good for anyone. (Delphion understands the problem and will at some point develop security access privileges for corporate customers. In the meantime, it mentions that Web site access products are available on the market.)

Even worse: Imagine middle and upper management saying, "Why should we pay a six-figure subscription to Derwent when we can search their fancy indexing on Delphion? Let's cancel our Derwent subscription!" This would be catastrophic for everyone. Derwent would lose revenue rather than increasing it; searchers would tear their hair out in frustration at having to cope with an unsuitable search platform; and search clients would lose out most of all from the quality of search results they would receive.
 

Motivation
Where is Delphion coming from? Delphion is trying to establish its visibility as an independent company, having left the aegis of IBM on May 15, 2000. To do so, it is introducing new functions and resources, and Derwent is certainly one of the most visible. Delphion wants to lay the foundation for future enhancements during the coming months. Delphion seems very motivated to do this, and in fact Derwent on Delphion may look rather different when you read this article than it did when I wrote it.

Where is Derwent coming from? It says it sees this as an experiment in making its indexing available to a wider audience its "toe in the water" to get some initial experience and solicit feedback. I suspect Derwent is also probably in some financial straits now. Every time one Derwent subscriber company buys another one (and there's been a lot of that going on in recent years), Derwent loses a subscriber. As its subscriber base shrinks, Derwent must find other sources of revenue.

I am all in favor of Derwent's attempting to expand its market, since increased revenue to Derwent from previously untapped sources would take a bit of pressure off current subscribers. However, I don't believe this is the way for Derwent to do it. Derwent has released its indexing to an end-user platform that is just too basic to handle the challenges of searching Derwent to its full potential and opened it to an end-user community who can't have a clue of how to use the codes effectively. Bottom line: Professional searchers won't search Derwent deep indexing on Delphion, and end users can't, and shouldn't, try.
 

Conclusion
I talked in some detail in my last "Better Mousetrap" column about current efforts underway by the online hosts to connect Internet and online patent resources. In that article I said, "Let's face it Internet search engines will never match online search engines, nor should they try to. And online linking capabilities will never match Internet linking capabilities, nor should they try to. The future lies in better communications between the two kinds of resources." I visualized a rosy future in which professional intermediaries would search online databases to best advantage, end users would follow up on these searches using Internet resources to best advantage, and online and Internet hosts would maximize communications between each other.

It can happen. I see a best-case scenario here in which professionals could do in-depth patent searches online (using multiple databases and all the bells and whistles of the online hosts) and download just Derwent accession numbers or patent numbers to send to their clients. Delphion could provide an easy interface for end users to import the search results into Derwent on Delphion (without having to key individual numbers) to view Derwent abstracts and take advantage of Delphion's excellent linking and imaging capabilities.

Make it so!
 

Suggestions to Delphion and Derwent

1. Derwent and Delphion: Take the chemical and polymer codes off Delphion. This is an end-user platform, and the codes are not end-user tools. However, do leave the Derwent classes and manual codes.

2. Derwent: Add other subject access more amenable to end users. You already have IPCs. Add ECLA codes (European Patent Office classifications) and U.S. patent classes. These are both dynamic classification systems for which hierarchies change regularly as technologies change, so you will have to plan for at least annual reloads of classification information. (Add them to your online database as well, while you're at it.)

3. Delphion: Add excellent support for these subject-access tools. Load scannable hierarchies of the manual codes, IPCs, and ECLA classes. Make it possible for users to click on specific codes or select ranges of codes and have them automatically entered as search terms. Give users an easy way to specify the Boolean logic they want to use between codes some of the codes will be synonyms and should be connected with OR logic; others will represent different sub-concepts in a search query and should be connected with AND logic.

4. Derwent: Change the pricing mechanism! A very high per-search charge will scare off everyone, professional searchers and end users alike, and it isn't justified for manual codes. (Besides, weren't you going to make the chemical manual codes accessible to non-subscribers one of these days soon?)

5. Delphion: Develop a more sophisticated search platform that will at least let searchers create and combine Boolean sets.

6. Delphion: Get serious about training. Even the manual codes require some training for effective use. So do all of the patent classification systems.

7. Derwent: Investigate better ways to increase your revenue by giving non-subscribers access to your deep indexing. The first thing you should do is to let experienced independent search brokers use the deep indexing search tools for non-subscriber clients (at a high online rate to make up for their non-subscriber status).

Nancy Lambert's e-mail address is nela@chevron.com.

Table of Contents Previous Issues Subscribe Now! ITI Home
© 2001